T-1777

Managing a patient's medical history should not mean hours of searching through scattered paperwork. Yet clinicians routinely face fragmented, out-of-network records in the form of scanned lab reports, medical PDFs, and faxes. Important diagnoses, medications, and allergies can be buried across documents, making it difficult and exhausting to establish a complete picture of a patient's health. While AI can summarize these records, trusting generated information is another challenge. A hallucinated or misinterpreted clinical fact could have serious consequences. This creates what we call "hallucination anxiety." LOCUS is a multi-tenant, AI-powered medical intake SaaS designed to address this trust gap. Rather than producing untraceable AI summaries, LOCUS introduces a "Citation-First" clinical intelligence approach. It ingests unstructured medical PDFs, extracts structured clinical facts, including diagnoses, medications, allergies, laboratory results, and vitals, and mathematically maps each extracted fact to its exact spatial coordinates (bounding boxes) in the original document. For example, when a clinician clicks on an extracted Ejection Fraction of 35%, LOCUS automatically navigates to the relevant page and draws a precise highlight box over the exact source text. Every extracted fact can therefore be visually verified against its original evidence. This transforms AI from a black-box guesser into a verifiable clinical assistant, allowing clinicians to review information with greater confidence rather than blindly trusting generated text. Trust and patient safety extend beyond extraction. LOCUS features a "Fast-Fail" triage gatekeeper that blocks wrong-patient uploads before they enter a patient's chart, helping prevent cross-patient data contamination. Its L2 Asynchronous Alert Scanner continuously monitors the patient timeline and flags potentially life-threatening drug-allergy conflicts in real time for clinical review. The platform combines a Zero-DevOps Next.js frontend with a secure Supabase backend enforcing strict Role-Based Access Control (RBAC). Its multi-tenant architecture supports secure access across users and clinical workflows. At its core, LOCUS aims to reduce the burden of manually navigating fragmented medical records while addressing one of the biggest barriers to clinical AI adoption: trust. Clinicians should not have to choose between the efficiency of AI and the confidence of verified medical evidence. By combining automated extraction with exact source-level spatial verification, LOCUS turns unstructured medical paperwork into actionable, traceable, and trustworthy clinical insights, helping clinicians focus more of their time on patients.