Healthcare AI and founder-led product engineering
Scrubs Co-Pilot
I founded and built an ambient clinical AI platform that turns clinician-patient conversations into structured medical documentation.
What I built
Owned the product and engineering work from clinical workflow discovery through application architecture, infrastructure, and production readiness.
Built the workflow across ambient listening, transcription, retrieval, structured note generation, and the interfaces clinicians use to review information.
Set technical direction while balancing product speed with data security, clinical workflow quality, and dependable delivery.
System at a glance
- Ambient audio enters a transcription workflow designed around clinical conversations.
- LLM and retrieval steps transform relevant context into structured documentation rather than a generic conversation transcript.
- The application layer connects voice, data, APIs, and clinician-facing workflows into one product experience.
- Production infrastructure supports secure delivery, operational visibility, and iteration with clinical feedback.
Key decisions
- Designed around clinician workflow validation, because an impressive model output is not enough if it does not fit the way care is delivered.
- Treated retrieval and structured generation as a product system with security and review needs, not as a model demo.
- Kept ownership end to end so product tradeoffs, infrastructure choices, and clinical feedback could move together.
How I approached reliability
- Used real healthcare-professional feedback to guide output quality and workflow decisions.
- Built security and production readiness into the roadmap rather than treating them as a post-MVP concern.
- Structured outputs made it practical to review and improve documentation workflows over time.
Outcome
The platform reduced clinician documentation time by about 80% by focusing the AI system on the practical work of turning conversations into usable medical documentation.
Technology focus
- OpenAI
- Whisper
- RAG
- Supabase
- pgvector
- React
- Node.js
- Twilio Voice
- Docker
- AWS