Relay

Listens to the Report
Gets Transcribed
Fill Out Missing Info
Dispatch Map
Shows Info Across Devices

What it does

Every second counts in an emergency. Right now, paramedics are managing a critical patient in the back of an ambulance while trying to relay information over a radio — and hospitals are preparing blindly. We built Relay. A paramedic speaks naturally — 'male patient, 60 years old, chest pain, shortness of breath' — and Relay automatically parses that into a structured patient record using an LLM, streaming it live to the hospital dashboard before the ambulance even arrives. On top of that, Relay streams real-time vitals — heart rate, blood oxygen, blood pressure — over WebSockets, so the ED can monitor the patient the entire way there. And with our live Toronto map, hospital staff can see exactly where every unit is and who's incoming. We're not replacing paramedics — we're giving them a superpower. Hands-free, real-time, structured data from ambulance to emergency department. That's Relay

How we built it

Voice Input & NLP Pipeline:

Browser-based Web Speech API for hands-free voice capture
OpenAI GPT-4o for natural language parsing into structured Ontario Ambulance Call Report (ACR) fields
Intelligent follow-up prompting to capture missing critical information based on chief complaint

Backend:

FastAPI (Python) for high-performance async API endpoints
WebSocket connections for real-time bidirectional data streaming
In-memory state management for rapid prototyping (with PostgreSQL-ready architecture)

Frontend:

Next.js 14 with React for the hospital dashboard
Real-time form updates with optimistic UI rendering
Interactive Toronto map showing live ambulance positions using Leaflet/Mapbox
Live vitals monitoring charts with streaming data visualization

Data Model:

Aligned with Ontario Ministry of Health ACR standards (Form 7530-4714)
CTAS triage scoring, OPQRST pain assessment, SAMPLE history
Structured procedure codes, medication tracking, and timestamps

Accomplishments that we're proud of

Sub-3-second voice-to-form latency: A paramedic speaks, and the hospital sees structured data almost instantly
Ontario ACR compliance: Our data model maps directly to the official provincial Ambulance Call Report, making it production-realistic
Live vitals streaming: Real WebSocket-powered vital sign monitoring that updates in real-time on the hospital dashboard
Smart follow-ups: The AI asks contextually relevant questions — chest pain patients get asked about radiation and nitro response; trauma patients get asked about mechanism of injury
Auto-generated narrative: Relay produces a professional medical narrative from the structured data, saving even more documentation time
Actually usable UX: Clean, dark-mode medical interface that doesn't overwhelm — designed for high-stress ED environments