🌐 Live: clerkcase.com
Interview AI patients. Submit your differential diagnosis. Get multi-factor feedback.
This repo is a public showcase. The application source is private.
ClerkCase is an AI-powered clinical reasoning simulator built for medical students. Students take histories from AI-generated patient personas, work through differentials, and submit diagnoses — getting structured feedback on their clinical reasoning, not just whether they got the right answer.
- AI patient personas — varied presentations across specialties, with realistic history-taking dynamics
- SOCRATES framework — structured pain assessment built into the interview flow
- 100-point multi-factor scoring — feedback covers history-taking technique, differential breadth, diagnostic accuracy, and clinical safety
- Daily challenges — fresh case every day, designed for short bursts of practice
- Leaderboards — community-wide ranking of cumulative scores
- Currently trialing with medical doctors providing case validation
| Layer | Stack |
|---|---|
| Frontend | Next.js, TypeScript |
| Backend | FastAPI (Python) |
| AI | Google Gemini |
| Infrastructure | Docker, CI/CD |
Built solo from domain acquisition to production deployment.
Live and functional, in active trial with practicing doctors.
Built and operated by Toby Lee. Part of the Thetas.ai family of AI education products.