# Global Codex Hackathon: [San Francisco]: Project Gallery

- **Event:** [Global Codex Hackathon: [San Francisco]](https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco)
- **When:** Thu, Apr 16 at 10:00 AM – 8:00 PM (PDT)
- **Where:** 1515 3rd Street, San Francisco, CA
- **Hosts:** [Cerebral Valley](https://cerebralvalley.ai/u/cv), [OpenAI](https://cerebralvalley.ai/u/openai)
- **Projects:** 30 (5 placed)
- **Page:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery

## Projects

### 1. Slalom - Jarvis

Control Plan + Harness engineering to democratize and scale Agentic Delegation across every codex user regardless of engineering or product background

- **Team:** [Will Thieme](https://cerebralvalley.ai/u/will_thieme)
- **Demo video:** https://app.box.com/s/b00vddcjewyvyq9xtn5xq5vr6ro8xos7
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=1

### 2. GuardGPT

This project demonstrates how governed AI work can happen safely on sensitive data. It starts with a database's role-based access, where the same workflow shows masked or unmasked PII depending on the user’s role. Before anything reaches the model, copied prompts are inspected for hidden non-printing characters and malicious prompt-injection content. The system preserves the safe part of the prompt, removes the dangerous segment, and shows both the original and scrubbed versions for transparency. When data is retrieved from the database, dangerous stored text is also detected and quarantined before it reaches the AI summary path. OpenAI Guardrails protect the model boundary, while the application traces suspicious database objects, query history, and API activity to reveal where PII exposure may still occur outside the governed workflow. Together, those features turn governance from a passive control into an active remediation experience.

- **Team:** [Rob Koch](https://cerebralvalley.ai/u/robcube), [Peter Griffin](https://cerebralvalley.ai/u/cadet), [Sreevidhya Sreedharan](https://cerebralvalley.ai/u/sree_14), [karthik mannava](https://cerebralvalley.ai/u/karthikmannava83), [Jordan Bredehoeft](https://cerebralvalley.ai/u/jordanbrede)
- **Demo video:** https://youtu.be/eBORDPew3EQ
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=2

### 3. Casper

CASPER is a tenant-scoped status and trust platform that redefines how enterprises view various product service health.
Instead of relying on a general global status, CASPER provides a tenant-specific source of truth for all purchased licensed products.by combining:
Availability (is the system working?)
Compliance (is it operating within policy?)
Attribution (who owns the issue?)

Key Features:
Layerd status view: Availability + Compliance + Error Budgeting
Tenant-specific dashboards (no global default)
Policy-aware compliance evaluation
Attribution engine (product vs tenant vs dependency)
Unified workspace for incidents, support, and history
AI-powered explanations and decision support using Codex

- **Team:** [Selva Dharani Thirumaran](https://cerebralvalley.ai/u/SelvaDharani), [Vidhya Boopalan](https://cerebralvalley.ai/u/CASPERVidhya), [Jeyakumar Ramachandran](https://cerebralvalley.ai/u/JEYAKUMAR)
- **Demo video:** https://drive.google.com/file/d/1QduPzUpGMf715YBTJDeMf9Hng-wThKdv/view
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=4

### 4. Section

Workflow audit agent to help employees identify and prioritize the most impactful ways AI can help them be more productive in their roles amnd the best AI capability (Custom GPT, Project, etc) to automate their workflow.

- **Team:** [Tom Harpaz](https://cerebralvalley.ai/u/tomharpazsection), [Bobby Isaacson](https://cerebralvalley.ai/u/bobbyisaacson)
- **Demo video:** https://drive.google.com/file/d/1pQEpW6ebGe8e4zNUJ1ESPR9izNDXZDkG/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=5

### 5. ClawX

We are building an MVP for a client need: an edge-based, self-optimizing supply chain control system that continuously monitors operations, makes real-time decisions, and recommends or triggers actions. The system demonstrates a complete closed-loop pipeline:

Vision → Reasoning → Orchestration → Prediction → Action

Key capabilities:

- Detects disruptions in real time using vision AI 
- Understands system behavior over time using AI
- Simulates multiple corrective actions using a digital twin
- Recommends the optimal action with measurable impact
- Operates as a fully local, always-on decision loop
- Controls processes through robotic execution

- **Team:** [Aravindhan Ramakrishnan](https://cerebralvalley.ai/u/Arvi), [Manav Modi](https://cerebralvalley.ai/u/manavmodi)
- **Demo video:** https://youtu.be/BNSsumVY5Oc
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=3

### 6. SpecForge

Project Name
SpecForge
Summary
SpecForge is an evidence-grounded multi-agent council built to solve a real 3GPP pain: when teams interpret complex specs, a single AI answer can miss conditions, invent assumptions, or produce implementation guidance that is unsafe to ship. SpecForge uses multiple specialist agents, evidence retrieval, and a judge workflow to produce more reliable, decision-grade outputs.
Project Description
Our starting point was a real 3GPP standards problem. Engineers working with telecom specs often deal with dense requirements, exceptions, cross-references, and vendor-specific ambiguity. In that environment, a single LLM response is not enough. It may sound confident, but it can still miss mandatory headers, mix models, flatten edge cases, or hide uncertainty.
SpecForge fixes that by turning AI into a council instead of a single answer engine. It brings together multiple specialist agents to debate the question, ground claims in evidence, and generate a final synthesis that is far more trustworthy. It also benchmarks Single LLM, Single LLM + Evidence, and Full Council + Jury side by side, then visually shows where the weaker approaches made mistakes and how the council corrected them.
While it was born from 3GPP and telecom standards pain, the same system also extends naturally to code review, RCA, sprint planning, and other high-stakes decision workflows.
Key Features
* Built around real 3GPP standards interpretation pain
* Multi-agent council with evidence-grounded debate
* Side-by-side comparison of Single LLM vs Council workflows
* Judge Board showing mistakes, contradicted assumptions, and corrections
* Rich final synthesis with markdown, tables, and structured reports
* Editable agent cards with per-agent model selection
* Reusable presets for telecom, code review, RCA, scrum, and more

- **Team:** [Venkataraman R](https://cerebralvalley.ai/u/veramasu), [Krishnamoorthy Duraisamy](https://cerebralvalley.ai/u/kduraisa_codex), [Anil Kumar Neeluru](https://cerebralvalley.ai/u/ANEEL), [VELMURUGAN KRISHNAN](https://cerebralvalley.ai/u/velmkris_codex)
- **Demo video:** https://drive.google.com/file/d/13C-e_YhK822xW3SSeaNyx1AxNzwvM1M7/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=6

### 7. CDW

Application Factory is a CDW-internal meta-builder, an AI-powered platform that builds other applications on demand for CDW teams. Instead of engineers hand-scaffolding each internal tool, a CDW user describes what they need in natural language, and the Factory specs, generates, tests, reviews, and ships a working app end-to-end grounded in CDW's stack, patterns, and governance.

Key features:
1. Natural-language intake that converts CDW business requirements into a structured app spec (entities, workflows, UI, integrations)
2. CDW-aware template library
3. Codex-powered generation of backend logic, frontend views, and integration glue
4. CDW sandbox deployment (currently on host)
5. Iterative refinement

- **Team:** [Datthesh Shenoy](https://cerebralvalley.ai/u/datthesh-shenoy), [Michael Davidson](https://cerebralvalley.ai/u/realityenigma)
- **Demo video:** https://www.youtube.com/watch?v=49e1IWYtcQI
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=7

### 8. Hypercode - deepsense.ai

HyperCode is a repository intelligence layer for coding agents. It builds a hypergraph from code structure, git history, semantic summaries, and Codex session memory so one natural-language query can return ranked, connected, implementation-relevant files faster and more completely than grep-only search.

- **Placement:** 1st Place
- **Team:** [Szymon Janowski](https://cerebralvalley.ai/u/szymon), [Maksymilian Operlejn](https://cerebralvalley.ai/u/MaxOperlejn)
- **Demo video:** https://www.loom.com/share/6b7ec27565424ce8b1ae206ad73b29d5
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=8

### 9. Codex-Astartes

Codex-Astartes is a security validation platform for web apps and APIs. It combines a FastAPI scanner (POST /scan) for quick URL checks with a deeper multi-agent assessment pipeline that analyzes runtime behavior, codebase signals, and policy constraints to produce evidence-backed security findings.

Key Features
OWASP-focused coverage (Top 10 categories) with categorized findings and remediation guidance.
Evidence-first workflow: profiling, surface mapping, static/runtime checks, correlation, and deduplication.
Built-in safety guardrails: scope controls, allowed domains, bounded probing, and non-destructive defaults.
Demo UX included: React frontend “Sample Explorer” for vulnerability sample endpoints and testable security scenarios.

- **Team:** [Dale Pedzinski](https://cerebralvalley.ai/u/pedzindm), [Kevin Zhang](https://cerebralvalley.ai/u/ruikevinzhang), [Charles Mendelson](https://cerebralvalley.ai/u/Carobert), [Atul Shah](https://cerebralvalley.ai/u/greenstreak)
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=9

### 10. Log2Flow AIAgents

Log2Flow is an AI-powered decision engine that transforms raw system and application logs into actionable business outcomes.
It goes beyond traditional process mining by reconstructing how processes actually run across systems and users. The platform ingests multi-format event logs, normalizes them, and uses specialized AI agents to perform process discovery, performance analysis, variant detection, and conformance audits.
A unified context layer connects these insights and feeds an Outcome Decision Agent, which ranks actions based on KPI impact and recommends next steps — automate, escalate, or review — with full traceability.
The platform includes interactive visualizations, outcome scorecards with before/after measurement, and a lightweight user interface for exploring process flows and insights.
Designed as a reusable and scalable solution, Log2Flow supports cross-industry adoption through modular components and pluggable industry profiles.
Built end-to-end using Codex, it demonstrates how AI can accelerate development while enabling real-world, decision-driven applications.

- **Team:** [Mini Burgula](https://cerebralvalley.ai/u/RealAIBuilder), [Prakash Babu](https://cerebralvalley.ai/u/Prax), [Haripriya Prakasam](https://cerebralvalley.ai/u/Agent_HP), [Mayank Mandowara](https://cerebralvalley.ai/u/MMHCL), [OmarEbnElKhattab Hosney](https://cerebralvalley.ai/u/omkamal)
- **Demo video:** https://youtu.be/07fLbB8MgvM
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=10

### 11. Cognizant-IDP

Intelligent Document Processing

- **Team:** [Harsha Chatriki](https://cerebralvalley.ai/u/HarshaC), [Vinoth Panneer Selvam](https://cerebralvalley.ai/u/VP1985)
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=16

### 12. Forge

Forge is a clean-room, Codex-built campaign orchestration workspace that turns a campaign brief into supervised, personalized outbound drafts. Key features include campaign creation, CSV/manual lead import, searchable lead attachment, prompt versioning, a combined prompt editor with final injected prompt preview, single-lead preview generation, bulk generation, validation and review queues, approval controls, and a send-ready delivery flow with mock/Resend support. It is designed as a reusable, self-hostable reference implementation rather than a one-off demo.

- **Team:** [Tejo Vinay Potti](https://cerebralvalley.ai/u/tejovinay), [Abhilash Subhash Sanap](https://cerebralvalley.ai/u/abhis24), [tharun kumar bandaru](https://cerebralvalley.ai/u/THARUN18500)
- **Demo video:** https://drive.google.com/file/d/1R2JrLdkhux3IR4t9OXwA21Juc7Q_LXnA/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=21

### 13. Billable Tokens

Adaptive Codex is a Python CLI on top of the OpenAI API that routes every coding turn to the cheapest viable GPT-5.4 tier while preserving conversation context across model switches. The result is a coding session with the cost discipline — cheap tasks go to cheap models, hard tasks still get the frontier model, and the router learns from what actually happened instead of what someone wrote in a regex. 

Key features:
1. Portable conversation state. The app owns raw API output items in SQLite and passes them forward as input[] on every call, so context survives model switches without previous_response_id — a turn handled by nano is visible to a later turn on the frontier model.
2. Plugin-scoped taxonomy and memory. Task families are registered per plugin via routing/taxonomy.py; each plugin gets its own FAISS index and MemoryStore implementation (Protocol-defined). Another team can drop in a plugin='testing' or plugin='sql' and reuse the whole loop without touching the core.
3. Smart Retrieval-augmented LM routing. Every prompt is embedded (text-embedding-3-small), SHA-256-hashed for dedup, and indexed in a per-plugin FAISS IndexIDMap2 wrapping IndexFlatIP (inner product = cosine similarity on normalized vectors). gpt-5.4-nano sees the top-k neighbors and returns a typed RoutingProposal with model, task_type, task_family, memory_key, reasoning, confidence — enforced via Pydantic AI structured output.
4. Graceful degradation at every layer. Embedding failure → zero vector. FAISS missing → no neighbors. LM failure → neighbor-majority fallback. The CLI never sees an exception from route() — an engineering team running this in production doesn't get paged because the classifier had a bad day.
5. Tool-item continuity. function_call / shell_call / local_shell_call / apply_patch_call items and their matching outputs are persisted verbatim, so tool context carries across model swaps as well.
6. Lightweight stack. Python CLI + SQLite (WAL) + FAISS + vanilla HTML/JS dashboard. No training pipeline, no orchestration framework, no platform migration.

- **Team:** [Richuang Lin](https://cerebralvalley.ai/u/richlin), [Brian Norris](https://cerebralvalley.ai/u/bnorris), [Marc Saouda](https://cerebralvalley.ai/u/marcsaouda), [Ali Rezaeian](https://cerebralvalley.ai/u/Alirzn), [rajprakash bale](https://cerebralvalley.ai/u/rajprakashbale)
- **Demo video:** https://youtu.be/1M6gtJtOY40
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=11

### 14. Tribe AI

MCP Factory is a system that turns a structured PRD into a working MCP server automatically. Instead of hand-building every connector or workflow integration, a user defines the MCP’s tools, runtime configuration, integrations, and behavior in a PRD, and MCP Factory generates a Go MCP server scaffold, fills in tool logic with Codex-assisted implementation, generates tests and docs, and verifies the result. Key features include PRD-to-code generation, a web control deck for running builds, and a canvas/workbench for shaping PRDs visually.

- **Team:** [Andrew Enns](https://cerebralvalley.ai/u/Andrew-enns), [Jeremy Kirshbaum](https://cerebralvalley.ai/u/jkirshbaum-tribe), [Chris Robbiano](https://cerebralvalley.ai/u/Chris_Robbiano)
- **Demo video:** https://youtu.be/ve1ppbm0LcA
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=12

### 15. Agent Syn Labs

Synthetic Agent Lab is an AI-powered simulation platform for testing customer strategy decisions before launch. Teams can define a scenario
  (pricing, campaign, product launch, retention, competitor response), generate synthetic customer/persona and market context, and run a multi-agent
  simulation to see likely reactions and tradeoffs.

  Key features:

  - Workflow-based scenario setup (pricing shift, campaign resonance, product launch, etc.)
  - Ontology and graph generation from prompts, files, and links
  - Synthetic profile generation for customers, brand, and competitors
  - Step-by-step simulation runtime with streamed outputs
  - LLM-driven agent decisions and LLM-led metric synthesis
  - Admin evaluation endpoints for ontology quality checks
  - FastAPI backend + Vite/React UI, containerized with Docker and Cloud Run deployment support

  Goal:

  - Provide directional, forward-looking decision support (not deterministic prediction) so teams can compare options faster and reduce strategy
    risk.

- **Team:** [Adel Abdalla](https://cerebralvalley.ai/u/adelabdalla)
- **Demo video:** https://youtu.be/cx0r_G6GrCk
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=13

### 16. The LCA Collective

A Codex-powered admin audit workflow that captures the current ChatGPT Enterprise configuration, reviews related help content, detects changes over time, and generates a clean, audit-ready source of truth for client teams.

- **Team:** [Christopher Spencer](https://cerebralvalley.ai/u/Cjspencer42)
- **Demo video:** https://app.arcade.software/share/zlLth7ZJIN6BUypjeCRa
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=18

### 17. Eliza

Customer CRM System that glues the business and FDE needs to complete proper client deal flow.

- **Team:** [Roman Wicky van Doyer](https://cerebralvalley.ai/u/romanwicky)
- **Demo video:** https://drive.google.com/file/d/1BNUnRVeddFbwKX5T6WE8BehJhIXFNVNZ/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=20

### 18. Modernize Minds

Modernization App helps teams modernize legacy software faster, safer, and with less manual effort. It analyzes repositories to identify current technology stacks, outdated dependencies, code quality issues, and security risks. Beyond acting as an upgrade utility, it also serves as a recommendation engine by suggesting alternative technologies and modernization paths that can reduce total cost of ownership (TCO). Based on repository analysis, target technologies, and migration criteria, it provides clear, actionable guidance to help teams reduce technical debt, lower risk, improve long-term maintainability, and modernize with greater confidence

- **Team:** [Sathish Rama Murthy](https://cerebralvalley.ai/u/SATS2020)
- **Demo video:** https://youtu.be/t2rCb5qJk_s
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=22

### 19. Fractal Analytics - Meridian

Meridian is a Codex-powered healthcare intelligence platform for building and using governed healthcare data assets. It combines a custom meridian-healthcare plugin and MCP server with Codex agents and skills to inspect, clean, validate, and publish reusable data products, then exposes them through a Vercel app with governed chat, role-based access, audit logging, observability, and feedback capture.

- **Team:** [Apoorv Shrivastava](https://cerebralvalley.ai/u/Bot007), [sowmith mandadi](https://cerebralvalley.ai/u/sowmith), [Khyati Sahu](https://cerebralvalley.ai/u/Khyati), [Natasha Lalwani](https://cerebralvalley.ai/u/natasha), [Shadab Azeem](https://cerebralvalley.ai/u/ashadab)
- **Demo video:** https://youtu.be/4qzteP0i2pY
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=14

### 20. KYC Fabric

KYC Intelligence Platform is an AI-native commercial account opening and continuous KYC intelligence system built to show how banks can move from manual, fragmented onboarding to real-time, explainable compliance operations. The platform combines an executive command center, AI agent operations center, KYC case explorer, continuous monitoring map, and governance/explainability console into one connected workflow. Key features include AI-assisted document review, beneficial ownership visualization, exception handling, live monitoring alerts, decision traceability, and human-in-the-loop governance.

- **Team:** [Gagan Goswami](https://cerebralvalley.ai/u/gagangoswami)
- **Demo video:** https://youtu.be/SQs6wKIqJZA
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=15

### 21. Accenture Team 1

Apps are becoming part of the AI user experience, and quality now depends on how the whole app behaves in the wild.

- **Team:** [Ty Shaw](https://cerebralvalley.ai/u/darkelm)
- **Demo video:** https://youtu.be/pZx_V5wNb-g
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=17

### 22. Blank Metal

Codex-powered crawler maps the full site architecture that shows the interactive sitemap/viewer that plays in real time. It shows UX and design issues based on the 10 usability heuristics by Nielsen Norman Group with a recommendation panel using Codex-generated fix suggestions.

- **Placement:** 4th Place
- **Team:** [Michelle Thorsell](https://cerebralvalley.ai/u/MThorsell)
- **Demo video:** https://drive.google.com/file/d/1-KoUYzYUiXB4vaDJGdyONZNXLZAmyuLG/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=19

### 23. Altimetrik

Sprintpilot is an AI engineering agent that turns a Jira ticket into production-ready code changes, validation artifacts, and an automated pull request. Triggered from Slack with /resolve-jira TICKET, it fetches Jira context, clones the GitHub repo, creates a working branch, analyzes the codebase, implements changes, runs dual evaluations, generates SQL and markdown artifacts, commits the work, opens a PR, and posts progress updates back to Slack. It is designed for enterprise engineering teams that need faster delivery without sacrificing reviewability, traceability, or governance.

- **Team:** [Mohan Kocherlakota](https://cerebralvalley.ai/u/mohan_koch98), [Yasmeen Sultana Ahmed](https://cerebralvalley.ai/u/Yasmeen), [Adam Horvath](https://cerebralvalley.ai/u/horadam), [Atharva Joshi](https://cerebralvalley.ai/u/ajoshi)
- **Demo video:** https://www.loom.com/share/024dddc50f8647d4b7856fbbd9f78e63
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=23

### 24. Capgemini Utility ETR Predictor

This project involves implementing an advanced ETR system to improve the accuracy and consistency of outage restoration predictions. Operating in a secure, segregated environment, the solution will integrate with enterprise systems via APIs and event interfaces. It will combine rule-based logic with AI models, using data such as historical outages, network topology, weather, AMI signals, and field inputs to generate and continuously update ETRs in near real time. The system will provide confidence intervals (P50/P80/P95), ensure full traceability, and deliver explainable predictions, while integrating with key systems like OMS, SCADA, GIS, WFM, and CRM.

- **Team:** [Jaya Srivastava](https://cerebralvalley.ai/u/Jaya_Srivastava)
- **Demo video:** https://etr-demo.vercel.app/demo-video
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=24

### 25. MeltForge

MeltForge is an LLM-driven content intelligence platform for media and entertainment teams working with large video libraries. It helps users ingest footage, extract structured metadata, surface compelling moments, themes, quotes, and emotional beats, and turn that analysis into derivative assets such as highlight reels, trailers, and teasers. Key features in the prototype include a library-first video browsing experience, AI-powered video analysis, project-aware asset workflows, static clip preview publishing for rapid review, and a closed-loop approach that can use engagement signals to improve future outputs.

- **Team:** [Ronin Sharma](https://cerebralvalley.ai/u/RoninS)
- **Demo video:** https://youtu.be/CN-lCt1_yWI
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=26

### 26. Artefact

A powerful DevOps accessibility agent embedded in the CI process

A multi-step AI agent scans front-end code, recommends fixes and updates code to improve accessibility across user groups.

- **Team:** [Andrew Ash](https://cerebralvalley.ai/u/andrew_ash), [Pedro Mangino](https://cerebralvalley.ai/u/pmangino)
- **Demo video:** https://drive.google.com/drive/folders/1VDMt7UDlEHGltmrZUHoY6sGdcrlL4qcH
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=25

### 27. Team Andela

Tanzu
Internet-as-a-Service, powered by community where people can rent reliable connections from nearby hosts or request mobile internet delivery on their workspace/home, helping developers stay productive despite unreliable connectivity. With the possibility to set alarms, so whenever an outage happens, a recurrent purchase that is active, gets used.

In many regions, especially across Africa, reliable internet is not consistently available at home or during outages. However, connectivity often exists in pockets  like certain homes, offices, or locations may have stable internet at specific times.

- **Placement:** 2nd Place
- **Team:** [Cory Hymel](https://cerebralvalley.ai/u/cory), [Barun Singh](https://cerebralvalley.ai/u/barunio), [Adelynne Chao](https://cerebralvalley.ai/u/Adelynne), [Will Marquardt](https://cerebralvalley.ai/u/willma), [Venkat Gali](https://cerebralvalley.ai/u/vgali)
- **Demo video:** https://drive.google.com/file/d/1CLYFe2tbnJ3QhjAmolC9vIQDItojxyhH/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=27

### 28. FlowForge

FlowForge is a live architecture copilot for meetings. As a team talks through a system design, it listens to the conversation, turns key entities and relationships into an evolving diagram in real time, shows speculative ideas before they are confirmed, restructures the graph when the conversation changes direction, and then finalizes the result for export to Miro. In our demo, a spoken support-ticket workflow becomes a clean async architecture with escalation and audit paths, all without manual canvas editing.

- **Placement:** 3rd Place
- **Team:** [Anton Rubisov](https://cerebralvalley.ai/u/rubi242)
- **Demo video:** https://youtu.be/1PqnQo8xbsg
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=29

### 29. Deal Canary

We built a lightweight, agent-driven sales lead interface designed specifically for top-of-funnel sales workflows. Instead of forcing early structure like traditional lead systems, our product embraces the chaotic, unstructured nature of early sales interactions and helps account executives move faster.

- **Placement:** 5th Place
- **Team:** [Aaron Sander](https://cerebralvalley.ai/u/ajsander), [Mei-Chi Chen](https://cerebralvalley.ai/u/Mei-Chi), [Jack Harrison](https://cerebralvalley.ai/u/jackharrison), [Jing Brewer](https://cerebralvalley.ai/u/JingB), [David Yurkanin](https://cerebralvalley.ai/u/Dave-Y)
- **Demo video:** https://youtu.be/flVjLhldxDM
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=28

### 30. PwC Retail

Retail Digital Coworker is an AI-powered decision-support platform for retail merchandising teams that helps them understand, optimize, and act on gross margin performance across pricing, promotions, cost, and vendor funding.
Instead of static dashboards, it functions as an intelligent operating layer that:
Diagnoses what’s happening (e.g., margin down, sales up but profit down)
Explains why it’s happening (price, cost, promo, mix, funding, etc.)
Recommends what to do next (adjust price, change promo, renegotiate cost, apply funding)
Enables scenario planning (“what if we change X?”)
Supports execution with human-in-the-loop workflows
It is built around persona-specific workflows (Buyer, Planner, PDC/Sourcing, Merch Ops, Store Leader), ensuring each role gets insights and actions tailored to their decisions.

- **Team:** [Rian Oosthuizen](https://cerebralvalley.ai/u/rianoosthuizen)
- **Demo video:** https://app.arcade.software/share/nmS88zBi3wKRos2aAl9i
- **Project:** https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery?project=30

---

Markdown version of https://cerebralvalley.ai/e/openai-hackathon-sanfrancisco/hackathon/gallery. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt