# The Future of Agentic AI in Healthcare - Abridge x Anthropic x Lightspeed: Project Gallery - **Event:** [The Future of Agentic AI in Healthcare - Abridge x Anthropic x Lightspeed](https://cerebralvalley.ai/e/abridge-hackathon) - **When:** Sat, Jul 18 at 9:00 AM – 10:00 PM (PDT) - **Where:** San Francisco, CA - **Hosts:** [Abridge](https://cerebralvalley.ai/u/abridge) - **Projects:** 114 (6 placed) - **Page:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery ## Projects ### 1. Abby Abby is a patient-state agent built around the Abridge workflow. After an ambient clinical note is created, Abby continues the conversation outside the visit: it reads transcript/FHIR context, securely verifies the patient, performs a focused specialty-specific check-in, extracts interval changes and red flags, and prepares an action-oriented brief for the clinician. For this demo, Abby uses the synthetic Abridge ambient/FHIR dataset to show a full loop: admin and provider setup, patient outreach, OTP-style identity verification, patient chat, structured FHIR-shaped patient-reported updates, safety/evaluation gates, and clinician approval before any write-back or scheduling action. The goal is to turn Abridge from a note-generation endpoint into the beginning of an ongoing care loop: what changed, what matters, what needs clinician attention, and what can be safely queued for follow-up. - **Team:** [Oliver Aalami](https://cerebralvalley.ai/u/OliverA) - **GitHub:** https://github.com/aalami5/abby - **Demo video:** https://www.loom.com/share/db5a25789b3c4c059b420e14dcfd598a - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=114 ### 2. ecqms_thebelt electronic clinical quality measures (eCQMs) reporting is a swashbuckiling journey eveyr hospital makes to get paid by CMS (Centers for Medicare and Medicaid Services). Every year the reporting has become increasingly challenging, what with CMS increasing the mandatory requirements year after year. Hospital quality personnel are over-burdened as is and this planning tool can help them with a clear start and set them up for successful submission. The tool takes the measure selection as input along with details about the hospital/CCN and provides a consolidated, consultative report with clear deadline-driven timelines, action items and measure specific recommendations and risk/gaps analysis. - **Team:** [Priyadarshini ravindran](https://cerebralvalley.ai/u/Curie_5) - **GitHub:** https://github.com/prp86/AI-Hackathon - **Demo video:** https://studio.youtube.com/video/gIEVKeGUoRI/edit - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=1 ### 3. Rakibul Hasan Contextual DDI Agent identifies Drug - Drug Interaction within EHR when provider submits a new medication order. It fetches patient current H&P, Medication, Lab order, Past History, identifies potential adverse effect, confirms using UpToDate, summarizes using LLM. - **Team:** [Rakibul Hasan](https://cerebralvalley.ai/u/Rakib_H) - **GitHub:** https://github.com/shonkhochil/ddi-contextual-agent - **Demo video:** https://youtu.be/bbXh9v1XHIY - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=2 ### 4. RT-Claude A Clinician in the loop radiation therapy planning agent powered by claude. The agent drives a real dose optimizer, proposes prescriptions based on RTOG/QUANTEC guidline, and iterates on the plan - while the clinician reviews a 3d Before/After comparison and approves, rejects or modifies each step. - **Team:** [kanghoun lee](https://cerebralvalley.ai/u/kanghounlee) - **GitHub:** https://github.com/nightandweather/rt-claude - **Demo video:** https://youtu.be/UiPYlMvml1Q - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=3 ### 5. ChartGuard AI scribes draft the note; someone still has to trust it before it hits the chart. Today that's an overloaded clinician eyeballing every line. ChartGuard is an independent judge that verifies every AI-generated note for grounding, safety, and completeness, then routes APPROVE / REVIEW / BLOCK. - **Team:** [Minakshi Mukherjee](https://cerebralvalley.ai/u/nori_84853) - **GitHub:** https://github.com/adaboostmm/clinical_judge - **Demo video:** https://youtu.be/EgQM418z_6k - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=4 ### 6. Consent MD PEACEful. Patient-based Empathic Approach to Consent and Education with EMR integration and clinical intelligence. - **Team:** [Vivek Mohan](https://cerebralvalley.ai/u/Surjun) - **GitHub:** https://github.com/VMSurjun/consentmd-poc - **Demo video:** https://youtu.be/UJScWV4p_t4 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=5 ### 7. Rare Disease Agent Rare disease patients wait 5 to 7 years for a diagnosis. Usually not because the answer is unknowable, but because the clues are scattered. The cardiologist sees a thick heart wall. The nephrologist sees protein in the urine. A rheumatologist sees joint pain and writes down fibromyalgia. Nobody puts them on the same page. Rare Agents reads a clinical note and does four things. Claude rewrites the symptoms in medical vocabulary, so "burning pain in his hands" becomes "acroparesthesia," because that's the word the databases actually know. Each term gets matched to an official HPO code. Those codes get scored against rare disease profiles. Then it says what to do: order this gene panel, watch for this, here are the trials recruiting now. The rule I built everything around is that Claude reads, but it never decides. Only the first step uses a model, and it's forbidden from naming a disease. The rest is lookups and arithmetic, so it can't invent a symptom, a disease, or a gene. I benchmarked it on 669 real published patients. Right answer first 72% of the time, top three 91%. A dumb baseline gets 44%. I also ran Abridge's 25 routine encounters through it, obesity and prediabetes and physicals, and got zero false alarms. That matters more, because a tool that cries wolf gets switched off in a week. Tools like Mendelian's MendelScan already scan health records for rare disease, but they read structured fields. "He's never really sweated" never becomes a billing code. It only exists in the note. So I read the conversation instead. It doesn't replace a geneticist. It gets the patient to one years earlier. - **Team:** [Seon Min Kim](https://cerebralvalley.ai/u/Esther59) - **GitHub:** https://github.com/EstherKim97/Rare-agent - **Demo video:** https://youtu.be/PlLh-BTLSPA - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=6 ### 8. Care Signals Care Signals: a care navigator for every patient Self-funded employers and health plans already pay care navigation teams to do this work: call around, compare prices, steer patients to lower-cost sites of service. It works. It's also human-staffed and rationed - a few high-touch cases at a time, available to a small fraction of patients, and never in the moment the order is signed. Everyone else walks out of the visit with no idea whether their scan costs $600 or $3,400. Half of commercially insured Americans are on high-deductible plans, so before the deductible is met, they pay the full negotiated rate. Faced with an unknown number, they defer necessary care - or they get it and receive a four-figure bill they never agreed to. The prices exist. CMS requires every hospital to publish machine-readable files of negotiated and cash rates. But those files run to hundreds of megabytes, every system uses a different schema, and no patient can open one. Care Signals gives every patient a navigator, automatically, at the moment of the order. From the ambient note and signed order, an agent parses the order to a CPT code, normalizes facilities' published price files, matches the patient's specific plan, computes real out-of-pocket against their deductible, and cites every dollar to its source line. Where a facility publishes no rate for a payer, it says so rather than guessing. For example in the demo, across four San Francisco facilities for a single lumbar MRI, patient out-of-pocket ranged from $586 to $3,387 - the same scan, both prices legally required to be public, neither visible when the order was signed. The agent then messages the patient the ranked options, answers their questions from the facility data using Claude, books the lower-cost site, and sends a calendar invite with a reminder. What a navigation team does over days, for a handful of patients, running in seconds for everyone. Imaging is the beachhead, not the ceiling. The same pipeline - order in, priced options out - extends to every shoppable procedure in the published files: elective surgery, endoscopy, labs, infusions. It extends to high-cost drugs too, where the same problem has a different data source: pharmacy pricing, discount cards, and manufacturer assistance programs, which patients navigate just as blindly and abandon just as often. Ambient documentation already captures the order and flags the care gap. A flagged gap isn't a closed one. This closes it. Every price is pulled from a published source and traceable to it. Patient data is synthetic. - **Team:** [Shawn Dimantha](https://cerebralvalley.ai/u/shawnd) - **GitHub:** https://github.com/shawndimantha/Care-Signals - **Demo video:** https://youtu.be/EZkhFU1iS5k - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=7 ### 9. Clinical Decision Support Composer Most clinical decision support (CDS) rules are written once, in the abstract, and only discover their edge cases after they're live and firing on real patients. This composer moves that reckoning before go-live: a clinician writes a rule in plain English, and the tool grills it with sharp corner cases — grounded in your patient panel — so the rule ships already stress-tested against the population it will run on. - **Team:** [Jiahao Xu](https://cerebralvalley.ai/u/miboo) - **GitHub:** https://github.com/jhxu0416/bridgineering - **Demo video:** https://youtu.be/U2JB2WMP0jA - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=8 ### 10. Recourse Physicians complete an average of 39 prior authorization requests per week, spending about 13 hours a week on the process, according to the AMA's 2024 Prior Authorization Physician Survey of 1,000 practicing physicians. 94% say prior authorization contributes to burnout, and 93% say it delays patient care. Meanwhile, KFF's analysis of CMS data found that 80.7% of Medicare Advantage prior authorization denials that were appealed in 2024 were fully or partially overturned, yet only 11.5% of denials are ever appealed at all. The evidence to win most appeals already exists. Almost nobody has the hours to go find it. Denial Appeal Agent closes that gap. It reads a denied insurance prior authorization request (the denial letter, the patient's chart, and the payer's own policy) and finds the evidence, scattered across all three documents, that defeats each denial reason. It then drafts a fully cited appeal letter for a human clinician to review and send. The core pattern behind most winnable denials: insurers often cite a rule while overlooking their own documented exception to it. A denial might claim there is no completed 3 month drug trial on record, while the chart shows the drug was stopped early for documented toxicity, which satisfies a specific policy exception the denial never mentions. Finding that requires cross referencing three unstructured documents by hand. The agent does it live, with every tool call visible in the UI. Built as a raw Anthropic SDK tool use loop, with no LangChain and no framework. Four core tools (parse denial, search chart, check policy clause or exception, draft appeal) that the model calls in whatever order it decides, not a hardcoded pipeline. On top of that, an adversarial reviewer agent using Claude Haiku critiques each draft like a skeptical payer reviewer and can force a revision before the letter is finalized. This is an evaluator optimizer loop, combined with model routing (Sonnet drafts, Haiku reviews) and parallel execution of independent tool calls. Tested against both a synthetic demo case and a real, unedited UnitedHealthcare policy PDF downloaded during the event, proving the search tooling generalizes beyond a cherry picked sample. No autonomous medical judgment involved. The system finds and cites evidence, and a human decides whether to send it. - **Team:** [Sriharshini Gubbala](https://cerebralvalley.ai/u/Sri_Gubbala) - **GitHub:** https://github.com/Sri-H-G/denial-appeal-agent - **Demo video:** https://drive.google.com/file/d/1foH8OqUgtD45JOaijkxe0vl21KOFLTld/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=9 ### 11. Team Alex ED Discharge Follow-Up Coordinator - **Team:** [Alex Walsh](https://cerebralvalley.ai/u/Alex_walsh) - **GitHub:** https://github.com/alexbwalsh-max/abridge-followup-agent - **Demo video:** https://youtu.be/huAZAnnFpFY - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=10 ### 12. Vised.ai The only API that legacy applications understand is screen / mouse and keyboard. We transcribe display screen to json text format what allows computer use agent to act quickly and at extreamly small cost. Filling a form like this cost 0.2 cents. This works on any interface surface - both browser and native apps. We use both DOM to read browser and AX tree to read native surfaces. Vised gives agents hands and eyes - allow to actuate on any software which can be operated with mouse and keyboard. We do not send screenshots and dont use vision models - just a cheap LLM like gemini 3.1 flash. - **Team:** [Konrad Sierzputowski](https://cerebralvalley.ai/u/Vised) - **GitHub:** https://github.com/conrader/abridge-hackaton-vised - **Demo video:** https://www.youtube.com/@visedai - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=11 ### 13. Tross Pre-charting workflow using a voice agent to call patients before appointments. The agent gathers context, summarizes the visit, and writes notes directly into the EMR, saving providers significant time. - **Team:** [Meet Shah](https://cerebralvalley.ai/u/mcs19) - **GitHub:** https://github.com/getayden/ab-hack - **Demo video:** https://www.tella.tv/video/meets-demo-ay4y - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=12 ### 14. Health ID Health ID is a patient-controlled, source-preserving health identity for clinical-trial discovery. It imports a real ClinicalTrials.gov study (NCT03734029), evaluates 10 synthetic Patient Graphs against 12 eligibility criteria derived from the study, and uses an Anthropic tool-calling agent to resolve missing evidence from a connected synthetic record while preserving the original source. Health ID maintains criterion-level provenance, routes the opportunity to the patient for explicit consent, shares only a verified eligibility proof with the research site, and blocks requests outside the approved scope. It does not make final enrollment decisions. Every candidate remains subject to formal screening by the research site. - **Team:** [Igor Eduardo](https://cerebralvalley.ai/u/Igor-eduardo) - **GitHub:** https://github.com/nomad-link-id/health-id - **Demo video:** https://youtu.be/o43bcc42mc8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=13 ### 15. Consilium.ai Consilium is agentic treatment-decision support for relapsed/refractory aggressive B-cell lymphoma that goes chart to room to recommendation to action. A signals agent extracts the intangibles from the ambient visit (with verbatim quotes); a reasoner ranks the committed regimen set through a deterministic clinical rule engine, weighing chart and room together and flagging defensible off-guideline choices; a verifier grounds every rationale in curated primary literature and surfaces per-plan attention flags, failing loud if it cannot verify. The clinician selects plans and a summarizer produces an operational to-do list (orders, consents, referrals, monitoring) to send to downstream workflow. A signal from the room can correct the chart. Live Vercel link: https://hackathon-proj-phi.vercel.app/ - **Team:** [Shalin Kothari](https://cerebralvalley.ai/u/Kotharishalin), [Xianglun Mao](https://cerebralvalley.ai/u/VincentXMao) - **GitHub:** https://github.com/VincentMao/HackathonProj - **Demo video:** https://www.youtube.com/watch?v=cTs_kDBeWS0 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=14 ### 16. Canary Alzheimer's drugs only work if you catch it early, but people aren't screened until symptoms are obvious -- so patients who would benefit most get missed. Canary catches them earlier by screening an EEG they already took, raising a risk flag, then reading the visit note to filter out false alarms before escalating anyone to a doctor. - **Team:** [James Mu](https://cerebralvalley.ai/u/realjamesmu), [Derek Mu](https://cerebralvalley.ai/u/dmu) - **GitHub:** https://github.com/jamesmu03/canary - **Demo video:** https://youtu.be/rgb0_X0KYKk - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=15 ### 17. The Second Read An ambient AI scribe treats the doctor-patient conversation as ground truth and writes it down. The Second Read is the verification layer that reads the note back against the rest of the chart before it is signed. The target market is post-acute care and skilled nursing, where documentation drives both PDPM reimbursement, discharge safety, and continued patient care. The tool lives inside the sign note action itself. A single citation-grounded Claude pass reconciles the draft against every PT, nursing, OT, and consult note the physician might have missed. Each contradicted sentence is flagged with verbatim, timestamped evidence, alongside a drafted correction that the physician approves or dismisses one flag at a time. Three properties make it safe to trust: 1. No quote, no finding. Every citation is re-verified in code against its source document. A finding that cannot produce its supporting quote is dropped before the physician ever sees it. 2. Silence over false alarms. It understands that CMS Section GG tasks are clinically distinct, so a difference in scoring across tasks is not treated as a contradiction. 3.It flags in both directions. When the physician is right and the chart is stale, it flags the record, not the physician. The scribe heard one conversation in one room. The Second Read reads everything else: the nurse at 10pm, the therapist at 2pm, and the buried consult note. That is where the patient's real story might be hidden. - **Team:** [Rahil Patel](https://cerebralvalley.ai/u/rahilpatel0402) - **GitHub:** https://github.com/rahilpatel0402/The-Second-Read - **Demo video:** https://www.youtube.com/watch?v=T-2hOScAVqU - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=16 ### 18. Relay Relay is an agentic operations platform for inpatient nursing: promises spoken at the bedside become tracked operational objects with an owner, schedule, and trigger, and six AI agents check them against the patient's live chart, proposing actions that a deterministic policy gate tiers and a charge nurse approves, with every step audited. It solves two failures of hospital wards: deterioration that's caught late because threshold alarms ignore the hours of decline before them, and spoken commitments that silently expire in documentation. - **Team:** [Luis Landeros](https://cerebralvalley.ai/u/Superluis) - **GitHub:** https://github.com/superluis0/Relay - **Demo video:** https://youtu.be/ANBbdHFRt4o - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=17 ### 19. HomeReady HomeReady - verify the home before discharge. The problem Nearly every discharge plan ends with some version of "discharge home once safe." But nobody can verify "safe" from the chart. Post-discharge falls are a leading driver of readmissions, and home-safety evaluations (CDC STEADI, HSSAT) require a clinician home visit that almost never happens. The chart stops at the encounter — the home it's sending the patient back to is a blind spot. What it does HomeReady turns the Abridge care plan into a guided home walkthrough. A caregiver walks the patient's home with an Iphone app, and an agent runs the visit: reading the encounter note, deciding what to verify, asking adaptive questions, and grading the home against the actual discharge plan. It's not a scripted checklist. Grounded in Monica's record, HomeReady decides what matters for her — osteoporosis, a prescribed walker, mobility — and the decisive finding is a measurement, not a vibe: the bathroom doorway is too narrow for her walker, so the discharge plan cannot work as written. That evidence travels back upstream and coordinates a fix — routing findings to OT, coverage barriers to social work, and orders to the care team — then updates the chart for the final discharge decision. How it's built "Dumb client, smart backend." The iPad only scans and streams; all intelligence runs in a FastAPI backend. Three perception layers run live during the walkthrough: RoomPlan LiDAR for exact geometry (real walker-clearance math), a Claude vision fast-pass on camera frames feeding the voice agent, and a deeper Claude pass grading STEADI/HSSAT findings. The ElevenLabs voice agent's LLM *is* the backend — one brain sees the chart and the camera and writes Riley's next sentence. Findings write back as draft FHIR Observations, ServiceRequests (DME with Medicare coverage flags), and Tasks. What's real Every model call in the demo happens live. There is deliberately no fabricated risk score — findings are graded against CDC STEADI / HSSAT with per-patient rationale, and every order and escalation is drafted, never auto-sent. Monica is a synthetic patient from Abridge's ambient-FHIR dataset. Abridge captures the encounter. HomeReady verifies the home. - **Placement:** 1st Place - **Team:** [Pranay Madan](https://cerebralvalley.ai/u/pistachio_pranay) - **GitHub:** https://github.com/pistachiopranay/homeready-abridge-hackathon - **Demo video:** https://www.youtube.com/watch?v=EY5Lbxpk-V8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=23 ### 20. Sentinel Sentinel — the missing order agent for clinical conversations After every visit, spoken plans leak out of the EHR: doses that never got updated, referrals that bounce, follow-ups nobody booked. Sentinel audits the record against the conversation itself — and fixes it. Built today on Claude: the agent extracts every verbal commitment with word-for-word verified quotes, reconciles them against FHIR, and classifies discrepancies as WRONG, INCOMPLETE, or MISSING. An adversarial auditor agent challenges every finding before a clinician sees it. Repairs are FHIR-shaped, clinician-approved, editable, undoable, and fully audit-logged. Abridge captures the plan. Sentinel makes sure the plan happens. - **Team:** [Alex Koshykov](https://cerebralvalley.ai/u/AlexKoshykov), [Vaibhav Satishkumar](https://cerebralvalley.ai/u/VS-Coder) - **GitHub:** https://github.com/Visual-Studio-Coder/abridge-hackathon - **Demo video:** https://www.youtube.com/watch?v=DyY64J94bWY - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=25 ### 21. RangerLoop - Raj Singh Half of imaging referrals never complete. RangerLoop is a closed-loop care-execution agent that finishes the job. It takes a signed imaging order, or an Abridge ambient encounter, and extracts the study and deadline, every field evidence-linked to the clinician's own words. Then it works the phones: it calls imaging centers one at a time until it finds a slot on or before the due date (striking out slots that are too late), then places a real, guardrailed voice call to the patient to verify identity and confirm - never disclosing clinical results - and books the appointment. It tracks each order ORDERED - SCHEDULED, and escalates safely to a human when it can't proceed. Claude Opus 4.8 is the visible orchestrator, deciding each next action via tool use with its reasoning streamed live; a thin deterministic policy layer validates every choice (the agent decides WHAT; the policy decides WHETHER), including a safety gate that blocks any booking without verified patient identity. Every record is FHIR-shaped (ServiceRequest / Patient / Task / Appointment), ready to write back to Epic or Oracle Health. Built entirely today; the voice/telephony stack is pre-existing infrastructure consumed over an HTTPS API only. - **Team:** [Raj Singh](https://cerebralvalley.ai/u/hrhraj) - **GitHub:** https://github.com/hrhraj/rangerloop - **Demo video:** https://www.loom.com/share/0d5d94ef03ba45988e6a5d908fa61913 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=30 ### 22. Skin Team Psoriasis Progression Monitor is an adversarial two-agent chart review. Agent A builds the case that a patient's psoriasis is progressing, searching notes and clinical photographs, with every quote verified verbatim against the source. Agent B, a nurse-auditor who never sees the chart, cross-examines that case against a clinical rubric and the one thing the chart can't show: independent pharmacy fill records. Only claims that survive the debate become a verdict. - **Team:** [Nathan Wei](https://cerebralvalley.ai/u/nathnwei), [Ben Tran](https://cerebralvalley.ai/u/btonthego) - **GitHub:** https://github.com/salabajr/psoriasis-tracker - **Demo video:** https://youtu.be/8hGdfR7tEic - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=36 ### 23. BoardX No patient should become clinically invisible because they are waiting for a bed. BoardX is a clinician-supervised continuity agent for admitted patients who are still physically in the emergency department — keeping their story current, bringing forward meaningful changes, and helping the care team close the loop safely. - **Placement:** Finalist - **Team:** [Vishnu Ravi](https://cerebralvalley.ai/u/vishnuravi), [austin schoeffler](https://cerebralvalley.ai/u/aschoeffler) - **GitHub:** https://github.com/vishnuravi/boardx - **Demo video:** https://youtu.be/2lu8cy4BiC0 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=42 ### 24. SwiftCode SwiftCODE is a hands-free, edge-native agent for code blue and rapid response events. During a resuscitation, clinicians already call out medications, shocks, rhythm checks, airway placement and other interventions, but a recorder nurse must still capture those events manually under intense pressure. Paper documentation can be incomplete and timing can be inaccurate. SwiftCode listens to the room, transcribes speech in real time, extracts and timestamps clinical events, and runs advisory ACLS timers for CPR cycles and epinephrine intervals. The critical loop runs locally using SwiftInference, so transcription, event logging and timers continue even when hospital connectivity fails. Deterministic rules handle clear callouts instantly, while a local language model resolves more complex speech. Low-confidence events require one-tap human confirmation. At ROSC or termination, Claude converts the verified event timeline into a professional code record and plus/delta debrief, with structured JSON and Markdown export for downstream documentation platforms. SwiftCODE does not diagnose rhythms or direct treatment. It is a human-in-the-loop documentation and protocol-timing support tool designed to reduce recorder burden, improve timestamp accuracy and create a usable record of one of healthcare’s most chaotic moments - **Team:** [Kendall Ananyi](https://cerebralvalley.ai/u/kananyi) - **GitHub:** https://github.com/camananyi/swiftcode - **Demo video:** https://youtu.be/3BalvNBX1fs - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=48 ### 25. EVIDENTIA Evidentia is a provenance-aware, longitudinal, clinical diagnosis hypothesis-generation and Bayesian ranking system with next-best clinical questioning. Data from clinical unstructured notes are extracted and normalized into Human Phenotype Ontology (HPO) terms and are reverse indexed search via the Monarch Initiative (symptoms-> disease ontology map) to generate a list of disease conditions. Reactome pathway look up establishes potential linkages that strengthen the claim of these constellation of symptoms, creating a ranking system per episode of care. As the notes progress over time, the weight of the evidence shifts accordingly, self-adapting to the greater clinical picture. - **Team:** [Patrick Damaso](https://cerebralvalley.ai/u/dataphysician) - **GitHub:** https://github.com/dataphysician/evidentia - **Demo video:** https://youtu.be/dQTozhoG6P4 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=49 ### 26. Twiage Twiage is an ambient AI agent team for ED triage: while the nurse talks to the patient and takes vitals, an orchestrator and four specialist agents (Scribe, Vitals, Acuity, Routing) select the right Epic template, fill it live with transcript-cited, confidence-scored values, propose an ESI acuity and routing, and leave the nurse with just a single review and accept task. AI assisted triage can be more consistent when humans are fatigued, and they can make the conversations more rich since the information will get captured without any additional work from the humans, saving tons of time and also potentially lives. - **Placement:** 3rd Place - **Team:** [Kairi Wright](https://cerebralvalley.ai/u/stat-guy), [Nupur Garg](https://cerebralvalley.ai/u/Nupur) - **GitHub:** https://github.com/stat-guy/twiage - **Demo video:** https://youtu.be/2yaYkhNP5Ig - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=51 ### 27. Iridium Abridge writes the note. ProofNote makes it safe to sign: every billing code is backed by a clickable spoken line or FHIR fact, challenged by a second agent, and resolved with the clinician before it ships. - **Team:** [Dhawal Modi](https://cerebralvalley.ai/u/DhawalModi) - **GitHub:** https://github.com/Dhawal-Modi/abridge_hackathon - **Demo video:** https://drive.google.com/file/d/13zJPRw2sFysVY14Liv97pxuKvr2Y5XAW/view?usp=drive_link - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=52 ### 28. Praxess Prior authorization delays care because the evidence needed for approval is scattered across the clinical conversation, note, FHIR record, payer policy, patient, and outside providers. Key facts get lost between what the patient says and what enters the chart. Teams discover those gaps after submission, when the payer requests more information or denies the case. Praxess is a world-model prototype for prior authorization. It maintains a living case state from the moment the patient enters the room, with source provenance attached to each claim. Claude structures evidence from the transcript and separates doctor and patient turns. Praxess then verifies quoted spans against their source before adding them to the case state. It tracks the difference between clinician-documented, conversation-enriched, patient-reported, verified, and unknown information. The decision engine rolls candidate actions forward and scores them using expected approval lift, information gain, time cost, and staff burden. A person approves consequential actions. Each new observation updates the state and changes the recommendation. Praxess records reviewed state-action transitions, creating the foundation for learned dynamics as authorization outcomes accumulate. The demo follows one authorization from encounter capture through evidence recovery, patient outreach, external record retrieval, packet generation, submission, denial, and appeal. If you want to check it out yourself, here's the link: https://praxess-production.up.railway.app/ - **Team:** [Maya Gerdes](https://cerebralvalley.ai/u/mayagerdes), [Abhay Lal](https://cerebralvalley.ai/u/abby_laal) - **GitHub:** https://github.com/mjmgerdes/praxess - **Demo video:** https://youtu.be/Bjih9Y_RLZw - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=53 ### 29. Neurotrax Neurotrax is an agentic AI system designed to make telehealth encounters a richer source of structured information about neurological function. This creates a team of ambient agents that can assess and trend neurodegenerative disease progression in a standardized, quantitative way. Instead of treating video visits only as conversations, it coordinates independent speech and facial analysis agents that identify technically usable moments, measure bounded audiovisual features, abstain when signal quality is poor, and assemble the resulting evidence for clinician review. - **Team:** [Logan Nye](https://cerebralvalley.ai/u/logannye) - **GitHub:** https://github.com/logannye/neurotrax - **Demo video:** https://youtu.be/mANYc26opmU - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=61 ### 30. NUDG - Clinical An AI Companion that nudges clinicians and helps navigates complex workflows with an advanced UI/UX. Nudge Advice is agentic AI specialized in medicine that gives physicians a live second opinion inside their existing EMR workflow — saving time with real-time feedback, easing clinician burnout, and helping improve clinical outcomes. - **Team:** [Pablo Zavala](https://cerebralvalley.ai/u/pablomzavala), [Santiago Enriquez](https://cerebralvalley.ai/u/santiago9107) - **GitHub:** https://github.com/pazare?tab=repositories - **Demo video:** https://drive.google.com/drive/folders/1JLUa1jAU-d4eq9VQ4BQnSrV4fZOzMNJf?usp=drive_link - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=62 ### 31. Pill Patrol Kapsule is an EHR-embedded medication-intelligence backend that turns a patient's FHIR chart into a single, prioritized, evidence-grounded set of regimen recommendations — the interactions, duplicate therapies, unsafe doses, and missing guideline-directed treatments a clinician would otherwise have to catch by hand. - **Team:** [Rishika Iytha sridhar](https://cerebralvalley.ai/u/Rishika_iytha), [Girum Mekonnen](https://cerebralvalley.ai/u/Girum) - **GitHub:** https://github.com/iytharishika/Med-error-agents.git - **Demo video:** https://youtu.be/ndnQSfg9_rY - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=18 ### 32. Pablo's Mono Team A live patient/trial matcher that finds which trials a patient is eligible for. If connected to an EHR, this would be a constant flow of eligible patient for sponsors/CROs, and for clinicians this can be an additional revenue engine. - **Team:** [Pablo Albrecht](https://cerebralvalley.ai/u/PabloSF) - **GitHub:** https://github.com/palbrecht1/hackathon-abridge-2026 - **Demo video:** https://share.maxcare.ai/S2PX6zZw - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=21 ### 33. Cogniqa A developmental diagnosis triggers an administrative cascade that healthcare has never automated. 47 days. I waited for my son's first therapy appointment after his ASD diagnosis. Early intervention referral. Prior authorization request. School district notification. Documentation gap review. Each with its own form, its own deadline, and its own way of quietly failing if nobody follows through. That work falls on a clinician with three minutes before their next patient — or disappears entirely. Families of neurodivergent children routinely wait months for care that should begin this week. Not because the therapists aren't there. Because nobody handles what comes next. PostDx handles what comes next. What it does PostDx takes a FHIR R4 Bundle from the diagnosing encounter and runs six agents over it - **Team:** [Meerim Samakova](https://cerebralvalley.ai/u/777) - **GitHub:** https://github.com/Meerisha/post-dx-agent - **Demo video:** https://youtu.be/dYv9_lis5LI?si=dTyOU9d3y5p9y_0z - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=24 ### 34. UCSF GI TEAM At UCSF every patient scheduled for a procedure needs a call from anesthesia, which often doesn't happen, and procedures get cancelled last minute. This app processes referrals, runs pre-anesthesia triage, and handles post-op follow-up in one smooth flow. It even has a roster-aware helper agent (think Claude website's Fin) to assist app users clinically. - **Team:** [Brooklyn Bach](https://cerebralvalley.ai/u/Brooklyn) - **GitHub:** https://github.com/gitbklynb/napguard-live - **Demo video:** https://youtu.be/8XgPIe0o2FA - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=19 ### 35. Bedside Bedside is an AI system that automates routine hospital-to-family communication while enforcing a hard boundary on what AI is allowed to deliver. The problem: nurses spend 30+ minutes a day fielding repetitive phone calls from patients' families asking for status updates, and families of hospitalized patients have no reliable way to get information without calling the unit. Bedside takes a short nurse dictation plus the patient's FHIR chart data and runs every fact through an AI classifier with three outcomes: safe facts are automatically texted to family members in their own language, facts requiring context trigger only a nurse-approved "a nurse will call you" notice, and serious findings lock the texting channel entirely, the system schedules a human phone call and gives the nurse a briefing of what the family already knows and has asked. Family members can text questions back: the system answers only from already-shared facts or verifiable events (like whether a doctor has visited), and routes all clinical questions to the care team using a fixed, unvarying reply. The result is that nurses stop making routine update calls, families get proactive information instead of chasing it, and bad news is never delivered by AI, only by a human voice. - **Team:** [Sierra Hillman](https://cerebralvalley.ai/u/SierraHillman), [Julia Lin](https://cerebralvalley.ai/u/Juliajoy) - **GitHub:** https://github.com/sierranoelle/bedside - **Demo video:** https://drive.google.com/file/d/1nawhdwEW1DPthrwMRIlceSottwxI9zk0/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=20 ### 36. Beacon Our goal is to optimize every aspect of the clinical trial workflow for clinician, patient, and pharma stakeholders by extracting insights from voice inputs to increase operational efficiency in enrollment, adverse event detection, and clinical workflow support. - **Team:** [Jae Joh](https://cerebralvalley.ai/u/medicaldork), [Holly Tang](https://cerebralvalley.ai/u/Holly99) - **GitHub:** https://github.com/developingmedicine/Hackathonia - **Demo video:** https://www.youtube.com/watch?v=rJR-7x9apmQ - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=22 ### 37. GOATnote In emergency medicine, volume negates luck. However no one gets volume in the events that kill but don't happen often. HALO is an open-source ED module for High Acuity, Low Occurrence events, built into the track board nurses already use. When a mass casualty hits: deterministic reverse triage frees beds in milliseconds, one-click AI triage turns a 15-second field note into an evidence-quoted SALT category, a Claude agent reconciles nameless patients against the chart panel (propose-only, human-confirmed), and a second agent audits the log against EMTALA. Plus a simulation lab where rare decisions have consequences — before they have patients. Fail-closed, citation-backed, synthetic data, 275 tests. - **Team:** [Brandon Dent](https://cerebralvalley.ai/u/GOATnote) - **GitHub:** https://github.com/GOATnote-Inc/HALO - **Demo video:** https://youtu.be/KsQ1Zi96V_Q?si=ESgcHMfU4G00mC5O - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=26 ### 38. WardFlow WardFlow is an AI clinical operations agent for inpatient care that bridges the gap between clinical decisions and clinal execution. It helps on-call physicians coordinate and deliver patient care by ensuring critical actions are taken on time, urgent patients receive timely attention, and discharge plans are complete efficiently while clinicians remain in complete control. Our agent intelligently synthesizes potential next steps from a ward run, asks for clinicians to confirm actions, then executes the action (ex. ordering lab results). It will then be re-invoked when the lab results return and decide what the next steps should be based on findings, local polciies, and its own intelligence. It will formulate next steps, and allow the operator to execute them or modify while presenting its evidence. - **Team:** [Saad J](https://cerebralvalley.ai/u/codermanz), [Hamza Ejaz](https://cerebralvalley.ai/u/Hejaz) - **GitHub:** https://github.com/codermanz/InPatientFlow - **Demo video:** https://youtu.be/Jkov5KhfWgc - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=31 ### 39. Agent Interstitium An owner for the post-discharge results that fall through the cracks — follow-up decision engine that switches antibiotic therapy by susceptibility, or escalates to the treating physician with a HIPAA-minimized page. - **Team:** [ann hui Ching](https://cerebralvalley.ai/u/Annhui) - **GitHub:** https://github.com/annhuii/interstitium-demo - **Demo video:** https://www.loom.com/share/f909583777594edba5630e5b32f5db78 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=35 ### 40. CuidaHome This app allows caregivers to speak and log patient’s care information, and generates a pre-visit summary, and then sends it through the patient portal itself to me, with a computer-use agent. - **Team:** [Tam Tran](https://cerebralvalley.ai/u/tpminhtam), [Tu Nguyen](https://cerebralvalley.ai/u/Tunguyen) - **Demo video:** https://drive.google.com/file/d/1v-hqer1PMSxI6G9U-c7RYMDTu11Ttpk4/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=39 ### 41. Gage & Hiro This is a wound care companion app to help patients follow and guide their wound care treatment. - **Team:** [Gage Caudell](https://cerebralvalley.ai/u/gagemanning) - **GitHub:** https://github.com/gagemanning/ProMend_Hackathon - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=74 ### 42. Speech Therapy Pre-Check — situational speech profiling for speech-language pathologists The clinic is a non-representative speaking situation: people who stutter often speak very differently with a clinician than on a phone call or ordering food — so SLPs assess a sample that doesn't reflect the disorder. Worse, standard ASR can't even capture the data: it erases, smooths, or garbles stuttered speech (we tested it — sound repetitions come back as the wrong words entirely). Pre-Check runs a pre-visit session: a patient talks with a calm voice agent about their speech (the baseline sample), then takes a simulated phone call — ordering a pizza — that elicits real-situation speech. A voice interface built for this audience: tap-to-finish turns, no VAD cutoffs, it never hangs up on you. Recordings run through a two-engine pipeline: CrisperWhisper for verbatim words + timestamps (blocks detected from inter-word gaps, repetitions from tokens) fused with LLM-Dys for dysfluency type classification, anchored per turn. The SLP gets a one-page profile: how speech changed across situations, playable timestamped evidence for every detected event, and a side-by-side of what standard ASR heard vs. what was actually said. - **Team:** [Kristiyan Vachev](https://cerebralvalley.ai/u/ChrisVachev) - **GitHub:** https://github.com/KristiyanVachev/abridge-speech-therapy - **Demo video:** https://youtu.be/2l4jnHoOa3w - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=75 ### 43. Cely Handling ESL patient intake and speeding up medical note review in between patients - **Team:** [Karl Jude Rojas](https://cerebralvalley.ai/u/Karlj), [Luke Hartman](https://cerebralvalley.ai/u/lukehartman) - **GitHub:** https://github.com/luhart/cely - **Demo video:** https://www.youtube.com/shorts/cSEHp3Ec-zk - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=27 ### 44. MicroRecovery Sentinel — an agentic system that detects clinician burnout before it hits, acts to reduce it, and verifies the load actually dropped. Sentinel treats the clinician as the patient. Burnout shows up in the data before it becomes reality: rising after-hours charting, growing note backlogs, no recovery days — then suppressed HRV and eroding sleep. Sentinel ingests both sides — EHR activity and schedules as the leading signal, wearable vitals as confirmation — and computes a single Load Index against each clinician's personal baseline. Fully computed, never hardcoded. It helps rather than informs: one recommended action at a time, quick wins with formula-computed score impact, and unlogged strain spikes detected from wearable telemetry that the clinician can tag in one tap. It acts: drafts the note backlog for sign-off, proposes recovery blocks, and escalates to leadership with consent. A Claude-powered voice copilot reasons over the clinician's full context — ask why the score moved, negotiate the schedule, send a coverage request. Privacy is the architecture: wearable and personal health data shape the score but never leave the clinician's view — the hospital sees only the composite score and workload signals it already owns. And Sentinel closes the loop: it re-checks the data days later and verifies the load actually dropped. - **Team:** [Rushendra Sidibomma](https://cerebralvalley.ai/u/Rushendra), [Patrycja Brzozowska](https://cerebralvalley.ai/u/pati) - **GitHub:** https://github.com/Rushendra10/sentinel - **Demo video:** https://youtu.be/tl6EsnQpxzk - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=50 ### 45. AllClear AllClear turns a colonoscopy After Visit Summary into a day-by-day Road to Procedure, timed prep steps, today-only check-ins, a physician-voiced chat/voice companion grounded in the patient’s own record, so patients show up ready and fewer procedures get cancelled for inadequate prep. This app aims to bridge this pain point before a procedure and aids in navigation for people who find it hard to follow the boat load of instructions and notifies and helps them commit to it till the day of procedure. We use colonoscopy as our demo example but it could work with a variety of procedures as well. - **Team:** [Hetansh Kevadia](https://cerebralvalley.ai/u/hk47) - **GitHub:** https://github.com/Hkayy47/FollowThrough - **Demo video:** https://drive.google.com/drive/folders/1FZEMbttGpLe6QWtrwyzyLASfB1CPQ5yW?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=55 ### 46. TrialFlow AI This project aims to create a clinical trial operational system. - **Team:** [Monil Wani](https://cerebralvalley.ai/u/monilw), [Mahendra Chaudhri](https://cerebralvalley.ai/u/Max0047) - **GitHub:** https://github.com/monil286/Abridge_Hackathon - **Demo video:** https://youtu.be/-sR29Z1E8Bs - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=56 ### 47. Team Intercept An EMR inbox triage system that routes patient messages using documented clinical context, not just the message itself — built for the Abridge × Anthropic × Lightspeed Hackathon. - **Team:** [Bharath R](https://cerebralvalley.ai/u/bharathr), [Apoorva Kolhatkar](https://cerebralvalley.ai/u/apokol) - **GitHub:** https://github.com/Apoorva2597/Intercept - **Demo video:** https://youtu.be/3zukCVMNRIk - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=82 ### 48. Mishaal Ali Trauma care is time-critical and coordination-heavy. TXA loses ~10% of its survival benefit for every 15 minutes of delay; every minute of delay in massive transfusion raises odds of death by 5%; and handoff communication failures contribute to roughly two-thirds of the most serious preventable hospital events. Trauma Co-pilot is a care coordination agent that listens from first EMS call to trauma bay departure while executing critical protocols, closing gaps in care, and completing handoffs to the inpatient team, with provider oversight. - **Team:** [Mishaal Ali](https://cerebralvalley.ai/u/Mishaal) - **GitHub:** https://github.com/mishaal-ali/trauma-copilot - **Demo video:** https://youtu.be/xQKLTFtPDLc - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=85 ### 49. Throughline A daily voice companion that helps patients and providers prepare for monthly medication follow ups, track progress, and make more informed treatment decisions. In less than 60 seconds, patients describe their day naturally while the app captures patterns in medication coverage, sleep, appetite, mood, symptoms, and side effects. Providers receive a clear summary of trends, safety flags, and a draft symptom questionnaire. - **Team:** [Arshia Moghaddam](https://cerebralvalley.ai/u/arshia) - **GitHub:** https://github.com/ZygimantasKoncius/abridge-hackathon - **Demo video:** https://www.loom.com/share/16ff45dc131d40bb9cc8ffdedd2b9544 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=97 ### 50. Mohamed Ghoweba, MD ANTITHRA is an end-to-end antithrombotic safety platform that ingests longitudinal chart data, uses agentic LLM extraction plus a deterministic 25+ rule engine to draft anticoagulation recommendations with cited guidelines, and routes them for clinician co-sign. Approved orders dispatch atomically to the patient (Portal/SMS/WhatsApp), care team (EMR inbox), and pharmacy (Surescripts e-Rx) — every fact cited, every rule versioned, every action co-signed. - **Team:** [Mohamed Ghoweba](https://cerebralvalley.ai/u/M_Ghoweba) - **GitHub:** https://github.com/Ghoweba1988/antithra-heart-care - **Demo video:** https://www.loom.com/share/1a23179833e44b5986d63372c6499b3b - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=28 ### 51. Recon Recon is a conversation-to-chart reconciliation agent for ambient clinical AI. It compares the patient conversation, AI-generated clinical note, and structured FHIR chart to identify clinically meaningful discrepancies before the note is finalized. Rather than editing the medical record, Recon provides evidence-backed clarification prompts, allowing clinicians to quickly verify potential inconsistencies while remaining in full control of the final documentation. - **Team:** [Britney Forsyth](https://cerebralvalley.ai/u/britneytf) - **GitHub:** https://github.com/britneyf/recon - **Demo video:** https://youtu.be/Me1FE-1bIcs - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=29 ### 52. Tricorder Our agentic model searches Head CT scans as they are scanned and applies a pre-trained model to detect potentially deadly brain bleeds and flags them as positive, moving them up the list to be read first by the radiologist while simultaneously sending alert messages to all involved parties (ER doctor, radiologist). It solves the critical problem of missed and delayed diagnosis of brain bleeds, which can results in catastrophic results for patients and multiple tens of millions of legal damages - **Team:** [Alexander Skinderev](https://cerebralvalley.ai/u/SashaSkinderev), [Faiz Francis](https://cerebralvalley.ai/u/FFrancis) - **GitHub:** https://github.com/SashaSkind/TricoderMax - **Demo video:** https://youtu.be/5bW-WBpHbQE - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=33 ### 53. Overturn Overturn is an agentic pipeline that turns denied prior-auth claims into evidence-chained appeals - citing the payer's own policy against its own denial, with a configurable autonomy dial deciding when a human steps in. - **Team:** [Sai krishna Vinnakota](https://cerebralvalley.ai/u/vinnakota) - **GitHub:** https://github.com/vvnsk/overturn - **Demo video:** https://youtu.be/To5S_R6UmT8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=37 ### 54. Baton Rx Every specialist visit changes the med list, and for independent and rural PCPs especially, the faxed consult note leaves them guessing who owns what while the patient keeps taking a drug nobody meant to continue. Baton reads those notes from inside the chart the PCP already uses, reconciles them against the patient's FHIR record into a live ownership ledger, flagging orphaned drugs, catching cross prescriber interactions no single EHR can see, and drafting the messages that fix them. - **Team:** [Bryce Piro](https://cerebralvalley.ai/u/brycepiro), [Vivek bhatt](https://cerebralvalley.ai/u/Vivekbhatt) - **GitHub:** https://github.com/piro3333/abridgehackathon - **Demo video:** https://www.loom.com/share/c12f2eb320194335bf50d45d256074cb - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=38 ### 55. Amplifier Health A plug-in AI integration agent for medical scribes that analyzes the patient’s voice during a visit to flag health conditions, like depression, anxiety, or PCOS  that the conversation text alone doesn’t reveal, often months before they show up in the chart. Synthesizes voice analysis and conversation for a more holistic patient analysis and embeds directly into the ambient scribe's capabilities. - **Placement:** Finalist - **Team:** [Camille Noufi](https://cerebralvalley.ai/u/camillenoufi), [Vijay Ravi](https://cerebralvalley.ai/u/vijaysumaravi) - **GitHub:** https://github.com/amplifier-health/clinical-integration-agent-demo - **Demo video:** https://drive.google.com/file/d/1idKlEM9bSfqwwSWfHfxE919d1ehws3P0/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=98 ### 56. ConcussionCompanion Patients feel symptoms precisely but can't name them clinically — so doctors get noisy, incomparable data from weeks-old recall. Concussion Companion closes the gap: patients describe their day by voice, and an agentic Symptom Manager annotates their words like a teacher grading an essay — mapping each phrase to Post-Concussion Symptom Scale / SNOMED terms, teaching the vocabulary as a side effect, and asking at most three function-anchored follow-ups. Clinicians open the next visit to five weeks of standardized 4-axis recovery data instead of "how have you been?" Concussion-first, built from lived experience, extensible to chronic care. - **Team:** [Weiwei Zhang](https://cerebralvalley.ai/u/weiwzhang) - **GitHub:** https://github.com/weiwzhang/concussion_companion - **Demo video:** https://youtu.be/QlqKaUpEqns - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=99 ### 57. Trauma Tracker A single tool that listens to a trauma resuscitation and does the paperwork in real time. As the team talks, it catalogues what it hears — interventions, spoken vitals, timestamps — then, live: computes trauma scores, tracks the accreditation/quality clocks (trauma activation response, airway, sepsis, STEMI), captures the charges, and determines whether the encounter qualifies for the trauma‐activation reimbursement. At the end it prints a 2‐page report: page 1 a nursing transcription sheet, page 2 the billing/compliance summary. One device, opens in a browser, no EHR hookup required. - **Team:** [Michael Giardina](https://cerebralvalley.ai/u/mikejgiardina), [Damian Mosher](https://cerebralvalley.ai/u/mawsher) - **GitHub:** https://github.com/mikejgiardina/ttprod - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=101 ### 58. Vigil ER Symptom Monitor ER Waiting Room Patient Symptom Monitoring Agent - **Team:** [Pranav Sanghvi](https://cerebralvalley.ai/u/pranavsanghvi), [Charumathi Raghu Subramanian](https://cerebralvalley.ai/u/Charu12) - **GitHub:** https://github.com/vigilhackathon/vigil/ - **Demo video:** https://www.loom.com/share/4cf45cfc16aa4b61977030c6139638f2 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=103 ### 59. sure for sure Sure for Sure turns clinical conversations and the patient's record into a focused work queue: ask the patient, reconcile the chart, or suppress an already-settled issue. It weighs how confidently the patient asserts each claim against what the record supports, catching treatment changes and record conflicts that a clean note can flatten — surfacing what needs a clinician's attention without adding alert fatigue. - **Team:** [Summer Han](https://cerebralvalley.ai/u/summer_saulting) - **GitHub:** https://github.com/hangoeun16/sure_for_sure - **Demo video:** https://youtu.be/1q-vl40PHhc - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=32 ### 60. the-agent-will-see-you-now Safety-verified acute stroke screening from an audiovisual neuro exam — for clinical use. A clinician (nurse, resident, or attending) performs a focused neurological exam normally; MedBridge captures the resulting video clip — moving images and audio — and its multimodal Evidence Extractor fuses Claude vision over frame sequences (facial asymmetry, arm drift over time, gaze tracking) with Whisper over the audio (speech clarity, naming, commands) into source-attributed evidence; the Clinical Planner works the NIH Stroke Scale one item at a time; the Safety Verifier refuses to call an exam "normal" without the evidence to support it and fires an emergency escalation on stroke red flags. Time-critical ✓ ("time is brain"), clinician-facing ✓ (the physician stays the accountable decision-maker), safer (a missed or over-claimed stroke finding is the liability the Verifier prevents) ✓ — grounded on the public-domain NINDS NIHSS videos, auto-segmented into 218 gold-labeled item clips. - **Team:** [brendan murphy](https://cerebralvalley.ai/u/Fully_Connected), [Smriti Singh](https://cerebralvalley.ai/u/smritisingh26) - **GitHub:** https://github.com/csbrendan/the-agent-will-see-you-now - **Demo video:** https://youtu.be/7NfvxBx_4Uc - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=40 ### 61. Aimen Tariq I built an AI multiagent app that reads clinical notes for clinical trials. It has a built in scientist agent that tracks patients and helps recruit the right patient population for a given trial, plus a second agent that double checks whether the assigned trial is actually the correct match. - **Team:** [Aimen Tariq](https://cerebralvalley.ai/u/AimenT) - **GitHub:** https://github.com/atariq-hub/trial-match-ai.git - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=59 ### 62. CareQ Patients at the end of a visit often ask “what q’s should I be asking now” and the right q at the right time can make all the difference. We propose an agentic patient-facing platform that can ambiently surface gaps from the patient's perspective, grounded on data from their EHR and insurance info. Our vision is to democratize access to a care navigator to all patients, so care delivery can be faster, smarter and safer. - **Team:** [Gopanandan Parthasarathy](https://cerebralvalley.ai/u/nandanpg), [Priyansh Pradhan](https://cerebralvalley.ai/u/priyanshpradhan) - **GitHub:** https://github.com/Priyanshbro/hackathon_abridge.git - **Demo video:** https://drive.google.com/file/d/1Ffd0w9okcavTIw1G0cq6WIqLWomJPVkV/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=63 ### 63. Vigil Vigil is an identity-aware patient monitoring and continuous re-triage agent. It begins with a voice or video intake, creates an initial patient risk profile, links the arriving patient to that profile, and then uses cameras and microphones to monitor changes over time. When Vigil finds evidence of deterioration, it re-assesses the patient's Emergency Severity Index (ESI), checks in with the patient when appropriate, calls the charge nurse, and documents the incident as a clinical note and FHIR record. - **Placement:** Finalist - **Team:** [pranav achar](https://cerebralvalley.ai/u/prnav), [Sahiel Bose](https://cerebralvalley.ai/u/sahielbose) - **GitHub:** https://github.com/PranavAchar01/Vigil-AnthropicxLightspeedxAbridge-Hackathon - **Demo video:** https://drive.google.com/drive/folders/1NuMrhCFw1mPxPK1F-_0g9izS0tOJouIo?usp=drive_link - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=34 ### 64. UCSF GI PreChart PreChart is a pre-charting verification agent for the office visit. It drafts a specialty-styled clinical note from the chart and the ambient conversation, and — where the two disagree — it doesn't just pick a side. Its founding assumption is what makes it different: the chart and the conversation are two fallible witnesses (charts go stale and carry copy-forward/coding errors; patients misremember, use lay terms, and get mis-transcribed), so PreChart trusts neither by default. For each claim it gathers adjacent, objective evidence, proposes which account the evidence supports, and surfaces the conflict for the clinician. The clinician can accept the agent's call or select the truth themselves — and the note updates live to match. Nothing high-stakes is auto-written; the clinician signs off. - **Team:** [Jin Ge](https://cerebralvalley.ai/u/egnij) - **GitHub:** https://github.com/jinge-ucsf/prechart - **Demo video:** https://youtu.be/wFKTnf4aTAE - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=43 ### 65. HORIZON HORIZON: Copilot for Non-clinical context - **Team:** [Vineet Tiruvadi](https://cerebralvalley.ai/u/virati), [Ruth Gebremedhin](https://cerebralvalley.ai/u/ruthgeb) - **GitHub:** https://github.com/virati/thething - **Demo video:** https://youtu.be/Za_SaovfRW4 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=60 ### 66. The Bald Informaticists An ED patient's chart is often longer than a novel — and it keeps rewriting itself while a physician is trying to read it before walking into the room. Physician Co-pilot puts an AI layer directly in the chart to fix that: an instant, rules-based clinical decision support engine flags what matters the moment new results land, with every recommendation grounded twice — the exact chart finding and the guideline behind it — plus one-click, pre-filled actions to act on it immediately. Underneath, a multi-agent LLM pipeline synthesizes a plain-language pre-visit brief and delta updates as the case evolves, personalized to each physician's preferences over time. In our demo, an ordinary chest-pain workup evolves live into a confirmed aortic dissection: the system catches it, orders the confirmatory scan, and hands the physician five ready-to-send actions the instant it's confirmed. Although this demo shows only one specialty, this can be architected to scale to all fifty specialties Abridge already supports across settings. Now that's scale. - **Team:** [Stephon Proctor](https://cerebralvalley.ai/u/Stephonomon), [John Lee](https://cerebralvalley.ai/u/Jsleemd) - **GitHub:** https://github.com/Stephonomon/abridge-previsit-copilot - **Demo video:** https://youtu.be/TSwHBSF12vo - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=71 ### 67. Asha: Autonomous Post-ER Referral Workflow Asha: Autonomous Post-ER Referral Workflow When a patient is discharged from the Emergency Room, the burden of coordinating follow-up care falls entirely on their shoulders. Nearly 50% of patients fail to schedule required follow-ups, leading to a 19% increase in 30-day ER readmissions. Meanwhile, Primary Care Physicians (PCPs) are bogged down by overflowing inboxes of routine, administrative referral requests that eat into complex diagnostic time. Asha is an autonomous, multi-agent voice workflow designed to eliminate this friction entirely. When a patient is discharged, an Intelligent Triage Engine (powered by Claude 3 Haiku) ingests the ER's FHIR discharge summary and evaluates the complexity of the required follow-up. If it identifies a routine, administrative task (e.g., a simple orthopedic referral clearance), it routes the case to our "Fast-Track" queue. From there, Asha proactively calls the patient to secure consent, calls the doctor's office to arrange the fast-track approval, and finally calls the patient back to confirm—zero manual phone calls required from the patient, and zero 30-minute diagnostic blocks wasted by the doctor. Crucially, Asha is built for healthcare compliance: it is not a "black box." The triage engine is heavily constrained to output its clinical reasoning as structured JSON, ensuring every automated routing decision leaves a transparent, clinician-auditable trace. Asha turns healthcare's most frustrating administrative bottleneck into a seamless, autonomous experience. Asha: Autonomous Post-ER Referral Workflow When a patient is discharged from the Emergency Room, the burden of coordinating follow-up care falls entirely on their shoulders. Nearly 50% of patients fail to schedule required follow-ups, leading to a 19% increase in 30-day ER readmissions. Meanwhile, Primary Care Physicians (PCPs) are bogged down by overflowing inboxes of routine, administrative referral requests that eat into complex diagnostic time. Asha is an autonomous, multi-agent voice workflow designed to eliminate this friction entirely. When a patient is discharged, an Intelligent Triage Engine (powered by Claude 3 Haiku) ingests the ER's FHIR discharge summary and evaluates the complexity of the required follow-up. If it identifies a routine, administrative task (e.g., a simple orthopedic referral clearance), it routes the case to our "Fast-Track" queue. From there, Asha proactively calls the patient to secure consent, calls the doctor's office to arrange the fast-track approval, and finally calls the patient back to confirm—zero manual phone calls required from the patient, and zero 30-minute diagnostic blocks wasted by the doctor. Crucially, Asha is built for healthcare compliance: it is not a "black box." The triage engine is heavily constrained to output its clinical reasoning as structured JSON, ensuring every automated routing decision leaves a transparent, clinician-auditable trace. Asha turns healthcare's most frustrating administrative bottleneck into a seamless, autonomous experience. - **Team:** [Madhur Garg](https://cerebralvalley.ai/u/RWEAI), [Saket Toshniwal](https://cerebralvalley.ai/u/saket) - **GitHub:** https://github.com/MG1100/asha-referral-agent - **Demo video:** https://share.descript.com/view/xuk9U2VUxnd - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=41 ### 68. INLOOP Discharge planning agent displaying on Patient Lists and Patient Charts - **Team:** [Caroline Luu](https://cerebralvalley.ai/u/carolineyluu), [Jhonatan Munoz](https://cerebralvalley.ai/u/MunozMD) - **GitHub:** https://github.com/carolineluu/inloop - **Demo video:** https://www.loom.com/share/ab67edc64dc640838f4c9ac1e47db371 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=44 ### 69. PreOp Navigator PreOp Navigator is an autonomous pre-operative agent that turns dense perioperative guidelines into an executable workflow, keeping clinicians in control while eliminating the coordination burden that causes surgical delays and preventable complications. It's not a checklist or a chatbot — it's a tireless coordinator that works a case for weeks, so the guideline stops living in a 90-page PDF and starts running itself. - **Team:** [Soumya Murag](https://cerebralvalley.ai/u/Soumya26), [ayushe nagpal](https://cerebralvalley.ai/u/ayushe) - **GitHub:** https://github.com/ayushenagpal/abridge-healthcare-hackathon - **Demo video:** https://youtu.be/pKsjCYDF_oU - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=54 ### 70. OncoOS AI OncoOS AI is an autonomous oncology co-pilot that lives on top of the EMR and turns a new-patient referral into a guideline-complete, tumor-board-ready workup — without the phone tag, faxes, and follow-up that normally stretch across weeks of a care team's time. The moment a urologist clicks "Refer to Oncology," a chain of specialized AI agents goes to work inside the chart. A Gap-Finding Agent reads the patient's record, classifies NCCN risk, and pinpoints exactly which investigations are still missing for a complete staging and molecular workup, citing the specific guideline page as verifiable evidence. It then closes those gaps autonomously: auto-filling and faxing the molecular (Caris NGS) requisition to the lab, and placing a real, natural-voice phone call to the patient that discloses it's automated, reassures them, and schedules their imaging. An Order-Entry Agent transcribes the call, extracts the agreed times, and books the orders back in the EMR. As results return, a multi-agent AI Tumor Board convenes and delivers a citation-grounded treatment recommendation with matched clinical trials, every claim linked to its source, or the agent abstains. The problem it solves: oncology workups are slow, fragmented, and frequently guideline-discordant, delaying time-to-treatment for the patients who can least afford it. OncoOS AI compresses the entire coordination burden, gap detection, ordering, patient outreach, scheduling, and expert review, into one autonomous workflow that surfaces right where the clinician already works. In the demo, a newly diagnosed high-risk prostate cancer patient goes from referral to a fully worked-up, BRCA2-driven, trial-matched treatment plan, hands-free. - **Team:** [roupen odabashian](https://cerebralvalley.ai/u/Roupen) - **GitHub:** https://github.com/Roupen92/OncoOSAI.git - **Demo video:** https://drive.google.com/drive/folders/1Z0kDcxFGi7HFFuhqmHim-ExFCjNIv7-e?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=76 ### 71. Safety Net Safety Net is a diagnostic-safety agent that re-reads an entire inpatient admission at the moment of discharge — every note, lab, result, medication change, and consult across all 14 days — and surfaces only the threads that genuinely fell through: an orphaned incidental finding, a home blood thinner held for a procedure and never restarted, a lab trend no single value reveals, a result left pending, a dropped consult recommendation. No clinician re-reads the whole chart at discharge; an agent can, with uniform attention, at exactly the decision point. Every finding carries a verbatim citation — an exact substring of the chart, validated automatically — and a specific, human-answerable question. It reasons about acknowledgment and equivalence, not keyword matching, so it suppresses look-alikes already closed in different words. It drafts the fix; a clinician approves; nothing sends itself. One engine, extensible to any record, handoff, or specialty. Patient now have a safety net for their health. - **Team:** [Kyle Walraven](https://cerebralvalley.ai/u/Kyle_Walraven), [Deep Nana](https://cerebralvalley.ai/u/DeepNana) - **GitHub:** https://github.com/kylewalraven-creator/Safety-Net - **Demo video:** https://www.loom.com/share/bab0226c060941509a32a2d72ab1b81f - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=45 ### 72. ByeByeHeadache 🧠 https://byebyeheadache.vercel.app/ ByeByeHeadache — the pediatric headache visit, understood in real time One in ten kids has recurring headaches. Most see a primary care pediatrician who has 15 minutes, a fragmented chart, and one question that actually matters: Is this a migraine, or something worse? ByeByeHeadache is a live visit-intelligence agent for exactly that encounter. As the doctor and family talk, an Anthropic Agent SDK pipeline listens to the transcript stream and builds the clinical picture in real time—headache phenotype, triggers, school impact, medication-overuse math shown with visible arithmetic, and a structured 14-item red-flag screen for secondary headache. Every fact links to its evidence: click any value and see the exact quote from the transcript or chart that supports it. If the agent doesn't have evidence, it says "unknown"—it never fills in the blanks. Prior history isn't retyped—it's pulled. ByeByeHeadache reads each patient from a Medplum FHIR server (patients, prior notes, PedMIDAS questionnaire responses seeded as real FHIR resources) and writes the visit summary back as a DocumentReference when the encounter ends. PedMIDAS—the validated pediatric migraine disability score—is captured live from conversation, scored when complete, and trended across visits. The demo ships four real scenarios: a first visit that ends with a scored PedMIDAS and a clinician-review plan; a preventive-medication plateau heading to neurology referral; a post-specialist follow-up improving on treatment (with interpreter turns in the transcript); and a red-flag escalation where the agent visibly pauses the routine pathway. Each visit exports two PDFs from one encounter: a clinical summary with trend charts for the doctor, and a plain-language action plan for the family. Everything is draft, evidence-linked, and clinician-reviewed. ByeByeHeadache doesn't diagnose or prescribe—it does the structuring, screening, and paperwork so the pediatrician can do the medicine. Stack: Anthropic Agent SDK (schema-validated extraction, grounded Q&A, insight generation) · Medplum FHIR (synthetic EHR, read + write-back) · FastAPI · React · ReportLab - **Team:** [Ramis Hasanli](https://cerebralvalley.ai/u/ramizik), [Aishwarya Taneja](https://cerebralvalley.ai/u/PedsMD) - **GitHub:** https://github.com/ramizik/abrdige - **Demo video:** https://youtu.be/ZNC4yZzV65o - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=46 ### 73. The Crown Our project is an adversarial auditor that verifies clinical notes against two independent sources of truth: the actual visit transcript and the patient's structured FHIR chart. It flags any claim that is not supported by either source and sends disputed claims to a panel of independent AI judges that surface disagreement instead of averaging it away. The result is a scored, evidence-cited safety report for every note. This matters because ambient scribing tools can automatically turn spoken conversations into permanent medical records, often without anyone checking the note word for word against what was actually said. A fabricated detail, a missed disclosure, or an overstated finding can then remain in a patient's chart without anyone noticing. When we tested our system on real, unmodified encounters, we found genuine errors that had gone unflagged, including a physical exam finding documented as performed when only imaging was discussed, a patient's age listed one year off from her actual birth date, a lab value labeled as "fasting" even though that word never appeared in the conversation, and a fabricated demographic detail with no supporting source. These examples showed us that an independent, evidence-based second pass can catch the kind of silent documentation errors that can put patient safety at risk before they ever become part of the medical record. - **Team:** [Arav Gupta](https://cerebralvalley.ai/u/zCranking), [Rajeev Sethuraman](https://cerebralvalley.ai/u/rajsethh) - **GitHub:** https://github.com/zCranking/Veritas-Hackathon-Project - **Demo video:** https://youtu.be/kHCMHPxuEnk - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=70 ### 74. Post Procedure Follow Up Team The problem: Post-hospitalization/post-procedure readmissions cost health systems billions a year and are penalized directly under CMS's Hospital Readmissions Reduction Program. Most of the early warning signs — worsening symptoms, medication confusion, uncontrolled pain — surface in the days right after discharge, in a phone check-in that may or may not happen, get documented inconsistently, or reach the right clinician in time. The idea: An agent places (or receives) a structured check-in call a few days after discharge, grounded in that specific patient's actual diagnosis, comorbidities, and medications. It scores the conversation against condition-specific red flags and routes the patient to one of four outcomes: urgent care referral, same-day physician callback, labs/imaging recommended, or routine follow-up — with a drafted chart note and a rationale a clinician can quickly verify. This is a natural extension of what Abridge already does (ambient conversation → structured clinical output), just pointed at a post-visit check-in instead of the exam room. - **Team:** [Lakshmi Subbaraj](https://cerebralvalley.ai/u/lsubbaraj), [Hannah Kim](https://cerebralvalley.ai/u/hankim) - **GitHub:** https://github.com/lakshmisubbaraj/post-discharge-triage-agent - **Demo video:** https://drive.google.com/file/d/1Wpoq0BjTw_dGc8r2-GSFwWWt3PTuMmgs/view - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=72 ### 75. TrialProof TrialProof is a human-supervised evidence-integrity auditor for clinical-trial pre-screening. Built with synthetic data, it uses Claude to extract exact chart evidence, then applies deterministic TypeScript rules to check protocol windows, thresholds, missingness, and semantic mismatch. The prototype flags unresolved criteria as Needs Data or Human Review, assigns next actions, and lets a coordinator add new evidence so only the affected criterion is reassessed with an audit trail. It does not determine final eligibility or enroll patients; it helps prevent incomplete evidence from becoming false certainty. - **Team:** [Kushal Naik](https://cerebralvalley.ai/u/drkushalnaik) - **GitHub:** https://trial-proof--drkushalnaik.replit.app - **Demo video:** https://youtu.be/04ovt6WzuLM - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=73 ### 76. BRIDGE by Aamir Javaid BRIDGE is an AI platform that bridges care between episodic hospital and clinic visits (20th century medicine) and continuous home-based monitoring (21st century medicine). With an initial focus on high-risk patients with chronic cardiovascular diseases (the top driver of morbidity/mortality and most expensive disease in the United States), the algorithm synthesizes insights from wearables -- such as vitals, weight, and single-lead ECG -- chat-based symptom assessment, and the patient's health record to detect disease earlier, notify the care team, and intervene to prevent hospitalizations and other adverse outcomes. This will result in improved health and lower healthcare costs. - **Team:** [Aamir Javaid](https://cerebralvalley.ai/u/aamirjavaidmd) - **GitHub:** https://github.com/aamirjavaidmd/bridge/tree/main - **Demo video:** https://youtu.be/SmBAonqa9XU - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=47 ### 77. AuthLens AuthLens is a point-of-capture prior authorization readiness agent that helps clinicians prevent avoidable denials before they happen. Prior authorization often fails not because a patient is ineligible for care, but because the evidence required to demonstrate medical necessity is missing, buried across the chart, or not clearly connected to the payer’s criteria. A clinician may make an appropriate treatment decision during the visit, yet administrative staff must later interpret a lengthy payer policy, search through notes and FHIR records, identify supporting evidence, request missing documentation, and manually assemble a submission packet. By that point, the encounter context has gone cold. The clinician may receive an inbox message days later asking them to clarify symptom duration, document failed treatments, rewrite the clinical rationale, or participate in a peer-to-peer review. This creates repeated interruptions, duplicated documentation, and administrative work that contributes directly to physician burnout. - **Team:** [Zubin Carvalho](https://cerebralvalley.ai/u/zubinc), [Pranesh Kumar](https://cerebralvalley.ai/u/pkumar1025) - **GitHub:** https://github.com/zubincarvalho/abridge-zubinpranesh - **Demo video:** https://drive.google.com/file/d/1GMH0v1dcudZmlO41S2DeXOLR6iNMHXHq/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=57 ### 78. AloraAI A case-manager co-pilot for behavioral-health patient transitions. Alora never places the call—the nurse does. Alora preps her, listens to the live call, grounds every answer in the approved packet, catches oversell, and learns from each outcome. - **Placement:** 2nd Place - **Team:** [Swindar Zhou](https://cerebralvalley.ai/u/swindar-zhou), [Hazel Wu](https://cerebralvalley.ai/u/hazelw) - **GitHub:** https://github.com/hazelwusy/AloraAI.git - **Demo video:** https://www.loom.com/share/a7451fb0e4f94890a0bcd6e14da283c5 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=58 ### 79. DualBeat DualBeat is an AI-powered cardiac intelligence platform designed to help clinicians evaluate the complete patient before making heart-related treatment decisions. The system combines cardiovascular data, multimorbidity analysis, medication-safety checks, clinical guidelines, ECG signals, CT imaging, and digital-twin simulation to identify risks that may be missed when care is focused on only one organ. The DualBeat logo represents two connected perspectives of cardiac care: the patient’s heart and the wider clinical context surrounding it. The dual-tone heart symbolizes the interaction between cardiovascular treatment and the rest of the body, while the central ECG waveform represents continuous analysis, clinical monitoring, and evidence-driven decision support. The orange and white color palette communicates energy, urgency, clarity, trust, and modern medical innovation. DualBeat does not autonomously diagnose or prescribe treatment. Instead, it provides clinicians with evidence-linked warnings, cross-organ medication conflicts, missing-information alerts, alternative options for consideration, and transparent simulation results. Its purpose is to support safer, more informed, and more personalized cardiac-care decisions. Tagline: Smart Care. Stronger Hearts. - **Team:** [Shradha Pujari](https://cerebralvalley.ai/u/ShradhaaaaP), [Chetas Parekh](https://cerebralvalley.ai/u/chetas) - **GitHub:** https://github.com/Shradhapujari/DualBeat - **Demo video:** https://drive.google.com/file/d/1mWarLuYWIVHE5ESh8yQqrIZIg6XHhqHo/view?usp=share_link - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=64 ### 80. Centific Clinical trials do not fail only because of science. They fail when the wrong sites take the wrong trials, with weak patient fit, poor economics, and high patient burden. WEI helps practices quickly decide whether a trial fits their patients, workflow, and business. As trials get more complex and recruitment costs rise, this is increasingly critical for CROs, sponsors, and physician practices. - **Team:** [Akshat Dasula](https://cerebralvalley.ai/u/AkshatDasula), [Prasanna Desikan](https://cerebralvalley.ai/u/prasannadk) - **GitHub:** https://github.com/AkshatMittu/clinical-trial-matching/tree/main - **Demo video:** https://www.loom.com/share/1c4775f17d1e4a8d84653d86a849240e - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=65 ### 81. Vetra Vetra is the AI-native operating system for veterinary clinics. It turns every incoming conversation into structured context, a supervised next step, and clear ownership, so more opportunities become completed care instead of getting lost between the phone, the front desk, and the clinical team. We start with triage and front-desk automation because that is where revenue leakage begins. Missed calls, poor routing, incomplete intake, and unowned callbacks can mean lost appointments, lost new clients, and missed follow-through. Vetra captures the request, surfaces urgency, applies the clinic’s rules, and gives the right person a clear next action. That is the wedge. The larger vision is an AI-native operating layer that makes each clinic’s protocols executable across its existing systems, from scheduling and records to follow-up, referrals, refills, and billing. - **Team:** [RISHAB Doshi](https://cerebralvalley.ai/u/Rish88), [Lakshraj Doshi](https://cerebralvalley.ai/u/lakshdoshi07) - **GitHub:** https://github.com/Rishabd2/vetra-ambient-scribe/tree/main - **Demo video:** https://www.loom.com/share/51053c06bb5946989bbd592e834f0ca7 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=67 ### 82. Sense SENSE is a consent-first agentic system for crisis prevention. During a stable moment, a user authors their own Stanley-Brown safety plan and approves exactly what the system may do during predictable high-risk windows like birthdays. Agents then execute that plan: a Companion Agent sends one evidence-based Caring Contacts message (brief, warm, non-demanding, the only messaging intervention with RCT support), while a Guardian agent validates every outbound action against the user's consent scope and cadence limits, blocking violations and writing an append-only audit trail. Not a chatbot, there is no chat. SENSE is an agent system whose headline feature is what it provably refuses to do. Built with the Anthropic API. Future work: Forecast, Circle, and Bridge agents per the repo README. - **Team:** [Leia Anapaula Chavarria-Davila](https://cerebralvalley.ai/u/leiaanapaula) - **GitHub:** https://github.com/LeiaAnapaula/sense - **Demo video:** https://www.loom.com/share/8596c49945154ac1a7d17a8420ab753b - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=69 ### 83. Salient Salient: the FDA knows the drug is dangerous. It doesn't know your patients. When the FDA issues a drug-safety warning, nothing re-reads the charts of patients already exposed. Prescribing alerts only check the next script; retroactive panel review is manual pharmacy work that takes weeks. Six months after the FDA's citalopram warning, 40% of exposed VA patients were still on high doses. Salient is an agent that closes the gap in seconds: Claude turns the alert prose into computable criteria, deterministic code scans the panel (the LLM never picks the cohort), Claude adjudicates the candidates using the chart and the ambient transcript, every claim is cited and mechanically verified, and a clinician approves the drafted action. Demo case: her chart says PRN opioids. Her transcript says "most days is the honest answer." She's nine weeks pregnant. The conversation caught what the chart missed. - **Team:** [Alex Cerullo](https://cerebralvalley.ai/u/AlexCerullo) - **GitHub:** https://github.com/AlexCerullo/Salient - **Demo video:** https://www.loom.com/share/53d501ce6e7b4e869f6f57ff84efa314 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=66 ### 84. Claimify Claimify is an intelligent pre-submission review workspace that helps healthcare teams produce cleaner claims by validating 837P data against coding rules, payer requirements, and the patient’s clinical record—before denials happen. - **Team:** [Akezhan Rakishev](https://cerebralvalley.ai/u/AkezhR), [semere Lemma](https://cerebralvalley.ai/u/RevCycleguy) - **GitHub:** https://github.com/Akezh/claimify - **Demo video:** https://www.loom.com/share/9872a4d6d2bb432e8f12fc0651108f04 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=80 ### 85. OpenCost Health Tool for patients that uses AI to extract billing codes from after visit summaries, provide cost estimate breakdowns, and search for cheapest facilities nearby to perform procedures. - **Team:** [Ben Goldberg](https://cerebralvalley.ai/u/benisgold), [Robbert Struyven](https://cerebralvalley.ai/u/rstruyven) - **GitHub:** https://github.com/benisgold/abridge-hackathon - **Demo video:** https://www.loom.com/share/17ebe1b696b34f29b8aea84fad7713c9?t=1 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=68 ### 86. Ruby Discharge Agent Ruby turns the conversations that already decide a discharge bedside and care-team rounds, into a reconciled, human-approved discharge plan. A Claude agent overrides the stale chart, catches the unspoken barriers behind readmissions, and updates the IDEAL checklist live. - **Team:** [Casey St Luce](https://cerebralvalley.ai/u/CaseyStluce), [Joan Kelly](https://cerebralvalley.ai/u/Caidir) - **GitHub:** https://github.com/casey1011/abridge-hackathon - **Demo video:** https://youtu.be/YoomjAfJe-g - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=77 ### 87. Code Clock Ambient, hands-free agent for resuscitations (code blue / code stroke). Our wedge is hands-free ambient extraction + real-time protocol reasoning + full source traceability. No taps. Ambient reasoning for the video below. - **Team:** [Mahtabin Rodela](https://cerebralvalley.ai/u/Rodela) - **GitHub:** https://github.com/AUW160150/for_Abridge - **Demo video:** https://youtu.be/NBjHDd3hU4M - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=78 ### 88. Faxclear FaxClear is an agent that reads the prior-authorization faxes drowning solo physicians, explains in plain English what the insurer actually wants, drafts the reply, and turns every fax into a deadline-tracked action item — so an approved MRI never silently dies because a vague 7-day form went unanswered. - **Team:** [Anupriya Inumella](https://cerebralvalley.ai/u/anupriya_i) - **GitHub:** https://github.com/Anupriya-Inumella/faxclear.git - **Demo video:** https://youtube.com/shorts/7dJkaDgsiVU?is=tQDiXg3_s38gWo6v - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=79 ### 89. PrplCare A surgical-episode agent for a clinic's patient-facing team. Golden path: **cataract** (the most frequent US surgery). Enrolled at referral, it preps personalized consult questions grounded in CMS volume/cost data; captures the patient's post-consult decisions (lens, cost, second-eye timing, med handling); coordinates the caregiver and confirms those decisions back to the clinic to schedule both eyes; then turns the discharge note into a day-by-day recovery plan and caregiver brief, runs drops/adherence check-ins, and escalates red flags through deterministic, tested code. Physician reviews save rubric rules the agent must follow in every future document. - **Team:** [Priyanshi Jain](https://cerebralvalley.ai/u/peajay), [Prakhar Jain](https://cerebralvalley.ai/u/kyuuurius) - **GitHub:** https://github.com/pea-jay/prplcare - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=87 ### 90. Enter your team name Type a cancer problem. The AI searches live oncology literature and public protein databases, then stocks the shelf with accession-backed proteins, NIH/NCBI 3D structure previews, source papers, and usefulness notes. - **Team:** [Nicholas Irving](https://cerebralvalley.ai/u/NickOne) - **GitHub:** https://github.com/NlCK01/HealthcareHackathon - **Demo video:** https://youtu.be/TlyuWF3eiFA - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=89 ### 91. Osti An open-source runtime authorization, workflow-resolution, and audit layer for clinical AI agents. Osti sits between an AI agent and clinical systems, checks every proposed action against patient context and organizational policy, routes legitimate exceptions to the correct human, and resumes only the exact action that person approves. - **Team:** [Ria Arora](https://cerebralvalley.ai/u/riaarora), [Dristi Chandra](https://cerebralvalley.ai/u/dristichandra) - **GitHub:** https://github.com/dristichandra/abridgehackathon-ria-drish.git - **Demo video:** https://www.loom.com/share/a66a86a3320543f39047421484a6690b - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=90 ### 92. Sentinel_Chronic_Agent Sentinel is a chronic health management agent between the visits. - **Team:** [Pavan Ghantasala](https://cerebralvalley.ai/u/PavanGhantasala), [Wiktoria Milczynska](https://cerebralvalley.ai/u/wikimilczynska) - **GitHub:** https://github.com/wikimcodes/Sentinel - **Demo video:** https://docs.google.com/videos/d/1NJBTnD2awmNH6oM8NxF94NPHJEM6hjnR5TL07Qs_gn8/edit?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=92 ### 93. Flow state Adaptive charting captures spoken bedside assessments in real time, converts them into structured clinical flowsheet entries, and verifies that flowsheet data is complete, correct, and current—while keeping the clinician in control. - **Team:** [Tunde Madandola](https://cerebralvalley.ai/u/Dimex08), [Nandita Damaraju](https://cerebralvalley.ai/u/NanditaD) - **GitHub:** https://github.com/Dimex08/Voice-to-flowsheet-quality-capture-agent - **Demo video:** https://docs.google.com/presentation/d/1rOdHi4QrwIhOdrlfHAD101rTXp-bCtow43jyCtrla0A/edit?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=100 ### 94. TRANSFER Transfer - a bed arbitrage agent that helps hospitals decant and offload patients who no longer need to be there by offering visibility to their ALC patients, and coordinate transfer to SNFs where need - **Team:** [Saad Ahmed](https://cerebralvalley.ai/u/theSaadMD) - **GitHub:** https://github.com/SaadAhmedMD/AbridgeHackathon - **Demo video:** https://youtu.be/8vKdFeGV8r8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=102 ### 95. StigmaGuard FHIR-native clinical documentation review using CDS Hooks to detect potentially stigmatizing language at note sign-off and provide explainable, context-aware recommendations while keeping clinicians in control. - **Team:** [Joseph Izzo](https://cerebralvalley.ai/u/izzomd) - **GitHub:** https://github.com/izzomd/stigma-detection-agent - **Demo video:** https://youtu.be/Yl9YOGgelX8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=81 ### 96. Priyanka RECALL — a pre-signature coverage agent that detects clinically important omissions in ambient-scribe clinical notes. While ambient documentation has largely solved hallucination, ~76% of its errors are omissions — content said in the visit that silently never reaches the note. RECALL audits every note before signature via an orchestrator–worker pipeline of separate agentic calls: (1) the transcript and encounter FHIR R4 resources are decomposed into atomic, typed clinical facts, each carrying provenance (verbatim quote or FHIR reference); (2) a grounded entailment judge marks each fact present/partial/absent against the note text alone, with verbatim evidence spans; (3) absent facts are classified for expectation and clinical severity; (4) a relevance filter surfaces only expected, safety-critical/major omissions, ranked, with an evaluator–optimizer patch loop that proposes minimal, evidence-grounded note diffs checked by an independent verifier. Because no ground truth exists for omissions, we manufacture it: an injection harness deletes known-present facts from gold notes (confirm-absent QC), yielding a fixed 69-deletion answer key plus 25 untouched false-positive controls — achieving 100% detection recall, a clean-note flag rate of 0.96/note after filtering (1.56 raw upper bound), 0.34% collateral instability, and zero invented facts in a source-vetted multi-agent audit. A stretch demo closes the loop by structuring transcript-only facts (e.g., an uncoded allergy) into FHIR resources and, with mandatory clinician approval, writing them to a sandbox HAPI server (not finished in time) - **Team:** [Priyanka Shrestha](https://cerebralvalley.ai/u/pri) - **GitHub:** https://github.com/Priyankas007/healthcare-agent - **Demo video:** https://youtu.be/ASmBSTbBhUg - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=86 ### 97. Prerna Jain + Kshtiji Aggarwal Industry data shows 40–60% of AI-flagged trial candidates get disqualified on first pass for reasons unrelated to clinical relevance, such as stale protocol versions, exclusion criteria (like medications) that never came up in conversation, and incompletely captured diagnoses. TrialGuard targets exactly those three failure modes. - **Team:** [Prerna Jain](https://cerebralvalley.ai/u/Prernaj), [Kshitij Aggarwal](https://cerebralvalley.ai/u/kshitij13) - **GitHub:** https://github.com/KshitijAggarwal/SecondReader# - **Demo video:** https://drive.google.com/file/d/1K9V_uCuWdZhKPF3M132pIfxe4byRvJRc/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=88 ### 98. Placer Most times in the hospital, when patients are ready for discharge, it takes multiple days to plan, driving up healthcare costs for Medicare patients by $24 billion a year. We built a product to automate that from day one, able to tackle over 400 discharge blockers. - **Team:** [Rahul Devathu](https://cerebralvalley.ai/u/devathu), [Ethan Schonfeld](https://cerebralvalley.ai/u/EthanSchonfeld) - **GitHub:** https://github.com/rdevathu/placer - **Demo video:** https://youtu.be/5_2w2EOcc9U - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=91 ### 99. Anesthesia AI Anesthesia AI provide on-demand info for anesthesiologists - **Team:** [Ben Ray](https://cerebralvalley.ai/u/Benmd) - **GitHub:** https://github.com/raymd23pro/Hackathon-Project - **Demo video:** https://youtu.be/jHAi4z4TJVQ - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=93 ### 100. Hospital digital twin Simulation Agent Based on the 25-patient clinical record provided, we built digital twins of the patients visiting an imaginary hospital. This prototype digital twin of an operating hospital can be used to model a real hospital patient flow, identify the areas of improvements needed, and test the efficacy of its implementation. Furthermore, by adding further patient data features, we can identify underlying patient-originated causes of hospital inefficiency to address and achieve operational improvements to serve them better. - **Team:** [Tatz I](https://cerebralvalley.ai/u/tatzi), [Georgia Liu](https://cerebralvalley.ai/u/georgial) - **GitHub:** https://github.com/georgia-max/synthetic-ambient-fhir-25 - **Demo video:** https://www.youtube.com/watch?v=j2Sk6OySfls - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=104 ### 101. dischargers An Abridge-powered agent that does the tedious prep for discharge medication reconciliation: it pre-populates a draft reconciled list from every source (home meds, inpatient orders, and the ambient transcript), then surfaces exactly where the sources disagree — each conflict shown side-by-side with the transcript quote and prescriber as evidence. The clinician still makes every call; we just turn "assemble the picture from scratch, then decide" into "the picture is already assembled and the disagreements are highlighted — decide." - **Team:** [akshay neema](https://cerebralvalley.ai/u/akshayneema), [Asit T](https://cerebralvalley.ai/u/asit98) - **GitHub:** https://github.com/asit2898/discharge-agent - **Demo video:** https://www.loom.com/share/9bd207eb9d0246a189beb02fa7a99222 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=83 ### 102. orchestra orchestra is a nursing-owned, ai-assisted clinical coordination platform designed to reduce delays and safety gaps during complex patient transitions. Nurses often hold the operational picture together across procedures, imaging, transfers, and discharges, yet they must coordinate through fragmented chart review, phone calls, and verbal updates. this makes it difficult for the bedside team to know whether every dependency has actually been completed. for this demo, this is an MRI readiness workflow for a critically ill pediatric patient. orchestra gives the bedside nurse a patient-specific readiness checklist, captures a natural clinical update, extracts operational facts, and maps them to readiness requirements in real time. it identifies confirmed items and unresolved dependencies such as transport staffing, equipment, consent, fasting status, interpreter access, anesthesia assessment, and cardiac-device planning. the ai does not medically clear the patient. It supports nursing coordination by extracting and reconciling information, while deterministic rules calculate operational readiness and maintain an auditable source for each update. MRI is the initial use case, but the same nursing-led orchestration engine could support OR handoffs, ICU transport, interfacility transfers, and high-risk discharge workflows. - **Team:** [janar t](https://cerebralvalley.ai/u/perdostoyevsky) - **GitHub:** https://github.com/imisshighfives/orchestra - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=94 ### 103. Consilium Generating treatment workflows based on multimodal patient context with up-to-date literature reviews. - **Team:** [Calvin Thomas Mathew](https://cerebralvalley.ai/u/calvinm), [Ahmad Sadek Dagher](https://cerebralvalley.ai/u/ahmad_dagher) - **GitHub:** https://github.com/calvinm7/abridge-hackathon - **Demo video:** https://youtu.be/0g4E5iAahSo - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=84 ### 104. Emmanuel and Ezra Bmt Agent - **Team:** [Emmanuel Asuzu](https://cerebralvalley.ai/u/Emman) - **GitHub:** https://github.com/Picannon/bmt-teach-back-agent - **Demo video:** https://youtube.com/shorts/6sGX9vn7R5c?is=NIShdsPM4L6f2Lw8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=95 ### 105. Yes You Health We’re addressing the data challenge in complex diseases like IBD, where many details matter. Today, significant friction prevents a clear picture of where each patient stands. Our workflow incentivizes patients to regularly inform and collaborate with their care team. A model of care maps disease activity, treatment stage, and potential next steps. With consent, structured data can support research despite access and Safe Harbor challenges. - **Team:** [Giles Pirio](https://cerebralvalley.ai/u/gip) - **GitHub:** https://github.com/gip/yesyouhealth - **Demo video:** https://studio.youtube.com/video/Im4izp_dZy/edit - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=96 ### 106. Contextual We are providing an AI-enabled means of adding contextual information to lab results (example here is an anemia workup) - **Team:** [Dr De](https://cerebralvalley.ai/u/drde), [Varun Satishkumar](https://cerebralvalley.ai/u/vsatish2017) - **GitHub:** https://github.com/varunsatish2017/Contestual.git - **Demo video:** https://www.youtube.com/watch?v=L8f99aCyVfY&feature=youtu.be - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=106 ### 107. Vyrelith Vyrelith is a connected triage and record-keeping system for people living with hidradenitis suppurativa (HS) — a chronic auto-inflammatory disease with an average diagnostic delay of 7–10 years, where most of the disease's story happens between clinic visits and never makes it into the chart. I built this based on my own experiences having Hidradenitis and ending up in the ER many times - trips that could be avoided. How it works. A patient texts the app the way they'd text a friend — typos, worry, and all. Claude (Opus, streaming) reads their message against their full FHIR record — conditions, medications, labs, the last clinic note, and the word-for-word visit transcript — plus every prior check-in, and does four things at once: replies in warm, plain language (built on HS-community language rules: "flares" and "tunnels," never "boils" or "contagious"), updates a structured symptom record, flags flares, oddities, and anomalies (an undocumented site, a silently stopped medication, a missed follow-up window), and notifies the right member of the care team directly — the patient is never told "call your doctor" as the plan; the system does the reaching out and says so. The clinician sees deltas, not transcripts. The console leads with what changed since the last update: before→after rows with the patient's verbatim words attached to each change, new and cleared flags, escalation shifts, and a prominent anomalies card — with a separate tab showing the agent's full reasoning trace (parse → chart → signals → criteria → route → safety), each step citing chart facts and patient quotes. The EHR itself is never mutated; check-ins live in an app-side record, provably leaving the FHIR source untouched. The data. Ships with 10 fully synthetic HS patients built for the demo — diverse in stage (Hurley I–III), scenario (ED abscess, biologic initiation, postpartum flare, HS+Crohn's overlap), and voice (minimizers, anxious question-askers, monosyllabic teens) — authored, adversarially audited, and fixed by a multi-agent pipeline, in the same FHIR R4 format as the hackathon's ambient-encounter dataset. - **Team:** [Amber Khan](https://cerebralvalley.ai/u/amberabbaskhan), [Ankit Maloo](https://cerebralvalley.ai/u/ankitm) - **GitHub:** https://github.com/amberabbaskhan123/abridgehack - **Demo video:** https://youtu.be/9FgkeZzKPBQ - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=105 ### 108. Spotted Zebra Spotted Zebra is a pediatric longitudinal clinical decision-support system that helps care teams identify growth concerns, unresolved specialist recommendations, and missed follow-up hidden across a child’s medical record. Children accumulate years of information across primary-care visits, specialist notes, growth measurements, referrals, appointments, scanned documents, and problem lists. The data may exist in the EHR, but the clinical story is fragmented. A pediatrician often has to manually search through old notes and charts to determine what was found, what was recommended, and whether anyone completed the next step. Spotted Zebra assembles these records into a longitudinal timeline and runs evidence-linked, deterministic clinical rules. During a pediatric visit, it can check whether expected growth measurements were collected, identify concerning growth trajectories, locate earlier specialist recommendations, and determine whether the recommended follow-up appears to have happened. - **Team:** [Srushti Sunil Madhure](https://cerebralvalley.ai/u/Srushti247) - **GitHub:** https://github.com/srushtismadhure/spotted_zebra - **Demo video:** https://www.loom.com/share/c70bdf7c168f406a82e30ec8303c1382 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=107 ### 109. MedTrack MedTrack is an interface to help healthcare providers, patients and caregivers track medication reconciliation and adherence from inpatient admission to discharge and post-discharge follow-up. A combination of messaging, LLM, and chatbot features ensures consistent planning throughout a patient’s hospital encounter. There are distinct epic and MyChart interfaces for the nurse navigator, primary care physician and patient and/or their delegate. - **Team:** [Seth Blumberg](https://cerebralvalley.ai/u/SethB) - **GitHub:** https://github.com/sblumberg/MedTrack - **Demo video:** https://youtu.be/q1eR4LQ45hE - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=109 ### 110. GreenLight Solves the clinician and Payer prior authorization faster, smarter AND safer. - **Team:** [Saran Teja Mallela](https://cerebralvalley.ai/u/nerdboss-stm) - **GitHub:** https://github.com/Nerdboss-stm/Greenlight - **Demo video:** https://youtu.be/Gs1GB-Ml1N8 - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=108 ### 111. Ju-Dol Our agent silently follows tumor board meetings, listens to the conversion while delving into the patients file. It runs a parallel multi agent board meeting and fills in the gap - what wasn't brought up during by the board, what new trials could the patient fit in, what (god forbid) mistakes were overlooked. It optimizes the next line of treatment for our patients and saves times for the physicians. unfortunately couldn't come up with a working demo video in time - filler video for now - **Team:** [Kian Dolatabadi](https://cerebralvalley.ai/u/kiandolatabadi), [James Hu](https://cerebralvalley.ai/u/jchu0) - **GitHub:** https://github.com/jchu0/tumor_board - **Demo video:** https://www.youtube.com/watch?v=lD3idp7Bb7U - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=110 ### 112. OncoMatch OncoMatch is an AI-powered clinical trial pre-screening agent that helps oncology care teams quickly identify and prioritize relevant clinical trials for a patient. Today, finding the right clinical trial is highly manual. A clinician or research coordinator may need to search through many studies and read long, complex eligibility criteria, then compare those requirements against scattered patient information such as diagnosis, biomarkers, treatment history, performance status, lab values, and imaging findings. OncoMatch automates this first-pass screening process. It takes a patient’s clinical context, searches for relevant oncology trials, converts complex eligibility criteria into structured rules, and evaluates the patient against those criteria using a hybrid AI + deterministic approach. Clear criteria such as age, ECOG status, and lab thresholds are evaluated with deterministic rules, including unit normalization. More nuanced criteria—such as biomarker interpretation, disease terminology, prior treatment context, or active versus treated brain metastases—are evaluated using Claude with supporting patient evidence. The system then ranks trials as High, Moderate, or Low Potential Match and explains exactly why each criterion passed, failed, is unknown, or needs clinical review. What problem does it solve? The core problem is: The right clinical trial may exist, but determining whether a specific patient could qualify is slow, complex, and difficult to scale. OncoMatch reduces the manual burden of reviewing trial criteria and helps care teams focus first on the most promising trials—without pretending to make the final eligibility decision. - **Team:** [Snehal Hattikal](https://cerebralvalley.ai/u/Snehal_Hattikal), [Sahithi Chimmula](https://cerebralvalley.ai/u/Sahithi_c) - **GitHub:** https://github.com/sahithic2000/oncology_clinical_trial_matching_agent/tree/main - **Demo video:** https://drive.google.com/file/d/1Y4BIIIVImmw81IpkdNueabc4KvBCZ-kV/view?usp=sharing - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=111 ### 113. The Gauntlet Before a patient leaves the hospital, 100 adversarial reviewer agents each get one job: find the documented way this discharge plan fails this patient — with verbatim chart receipts — or be refuted in review. A strict four-test judge (evidence, addressed, materiality, actionability; refute by default) rejects ~95% of objections on screen, with reasons: alert fatigue is filtered at the review gate, never forwarded to the physician. The few surviving findings render as evidence cards anchored to the exact chart lines that prove them, each with one staged order. The clinician accepts or dismisses-with-documented-rationale; accepted orders amend the plan; a targeted re-verification challenges only the amended dimensions: "0 surviving objections — attacked and not broken. Not a safety certification." Every receipt is machine-verified verbatim in code, so fabricated evidence cannot render, and every number on screen is measured, never simulated. The same swarm reviews any encounter from the organizer's synthetic-ambient-fhir-25 dataset (or a pasted/edited discharge summary), and every run exports CDS Hooks cards — the exact payload an Epic order-sign hook would render at discharge signing. Haiku 4.5 reviewer swarm + Opus 4.8 judge over a prompt-cached chart; ~$2 and ~2.5 minutes per 100-reviewer run. Built entirely today: pipeline, judge, CLI, and the EMR-style mission log, all rendering one live event stream. - **Team:** [Aseem Kumar](https://cerebralvalley.ai/u/aseemk) - **GitHub:** https://github.com/aseemk-15/gauntlet - **Demo video:** https://youtu.be/UfMsYBivzQc - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=112 ### 114. RxRescue RxRescue is an AI-powered patient navigation agent that helps patients overcome medication access barriers such as insurance denials, prior authorizations, affordability, manufacturer assistance enrollment, and pharmacy issues. Patients simply describe their situation in natural language. Using Anthropic Claude, the agent asks intelligent follow-up questions, determines the likely barrier, guides the patient through manufacturer assistance eligibility (demonstrated using the Pfizer Nurtec® PAP), and generates a structured summary that can be shared with the prescribing clinic. The goal is to reduce treatment delays, decrease administrative burden on healthcare teams, and help patients obtain prescribed medications faster while keeping clinicians informed. - **Team:** [Umee Davae](https://cerebralvalley.ai/u/DrUmeeDee) - **GitHub:** https://github.com/goodie3shoes/RxRescue - **Demo video:** https://youtu.be/TC9hQTc1XyU?si=kSLrvG5zZF7QE3qK - **Project:** https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery?project=113 --- Markdown version of https://cerebralvalley.ai/e/abridge-hackathon/hackathon/gallery. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt