The founder had spent eight years as a hospitalist before leaving clinical medicine to solve a problem she lived with every day. At most small and mid-size US hospitals, clinical coordinators were managing patient handoffs, bed assignments, and discharge paperwork entirely manually — whiteboards, spreadsheets, disconnected legacy software. The result: missed handoffs, delayed discharges, staff burnout, worse patient outcomes.
She had validated the problem with 12 clinical coordinators across 3 hospital networks, secured a letter of intent from an early partner, and earned a spot in Y Combinator's Winter 2025 batch. What she didn't have was an engineering team — or time. Demo Day was 10 weeks away.
Think like a product team, build like an engineering studio
The first thing we told the founder: don't build everything. With 7 weeks to a working demo, the biggest risk wasn't technical — it was scope. We ran a one-day product sprint in week 1 to identify the single highest-value workflow. The answer was patient handoff coordination — the thing every coordinator did 15–20 times a day and the thing most likely to go wrong without a system.
In week 1, we shadowed two clinical coordinators at a partner hospital for half a day each. We mapped the exact steps in a patient handoff: what information was needed, where it came from, who needed to act, and what the failure modes were. This session was worth two weeks of spec-writing.
What we built
Weeks 2–4 covered the core platform across three parallel tracks: the frontend coordinator dashboard, the backend API and real-time event system, and the HL7 FHIR integration layer. Stack: Next.js 14 with App Router, Spring Boot REST API with WebSocket server, PostgreSQL with event sourcing for audit trail, GCP Cloud Run for HIPAA-eligible infrastructure. Dr. Mehta was reviewing real screens by day 10.
Weeks 5–6 added the AI layer — the differentiator. We integrated an AI layer that analysed patient data and real-time bed occupancy to surface proactive recommendations: flagging patients at elevated discharge risk, auto-drafting handoff notes from structured EMR data. The AI handoff note feature alone reduced note-writing time from 8 minutes to under 90 seconds. In a 50-bed unit with 15 handoffs per shift, that's over 1.5 hours saved per coordinator per day.
Week 7 was about making it real. We ran a live 5-day pilot at the partner hospital with 3 coordinators using the system alongside their existing tools. We fixed 11 bugs, added 3 features coordinators asked for on day 1, and built the live demo environment used on Demo Day — a realistic simulation with 20 mock patients, live real-time updates, and all AI features running end-to-end.
Demo Day. Then the term sheets.
The founder demoed at YC's Winter 2025 Demo Day to an audience of 11+ investors. The live demo ran on the actual product — not a prototype, not a Figma mock, not a video. Real data, real AI, real product. Within 48 hours she had 5 investor meetings scheduled.
"The team didn't just write code — they acted like co-founders for 7 weeks. They pushed back when my scope was too wide, suggested the AI features I hadn't even asked for, and delivered a product I was genuinely proud to demo to 5 investors."
— Founder & CEO, YC W25 Healthtech Startup
