Synapse Responder
An ambient scribe that drafts EMS patient care reports while the crew works, and never stores a byte of patient audio.
- Draper Pitch Competition 2026 — Semifinalist, top 6 of 118
Every EMS call ends the same way: the patient is handed off, and then somebody sits in the back of the truck for twenty minutes typing up what just happened. I have done it. It is the worst part of the job, and it happens while the next call is already waiting.
Synapse listens to the scene through the device's own microphone, transcribes it, and feeds a live patient dashboard that updates during the encounter. When the call ends, the report is already drafted.
The architectural decision I am proudest of is the one that looks like a limitation: there is no backend and no database. Nothing persists, so there is no protected health information at rest and no HIPAA exposure to manage. That is privacy by design rather than by policy, and it is why the thing can actually be deployed.
What it does
- Shipped a browser-based ambient scribe that transcribes on-scene speech and auto-drafts EMS Patient Care Reports, cutting roughly 20 minutes of documentation per patient encounter.
- Built a live patient dashboard that updates during the encounter, not just a transcript after it.
- Architected the system deliberately without a backend or database so no patient audio or PHI is ever persisted.
- Semifinalist, 2026 Draper Pitch Competition, top 6 of 118 undergraduate teams.
Built with
- Node.js
- GPT-5.5
- Azure AI Foundry
- Web Audio
- Speech-to-text