AI-powered clinical training simulator: practice patient encounters with simulated cases, graded against published clinical guidelines.
Medkit is a gamified medical training simulator built by Bedirhan Keskin, a physician-turned-software-engineer from Istanbul. Medical students take medical histories, order labs, read imaging, diagnose and prescribe — all with simulated patients — and at the end of each encounter an agentic grader scores the whole session against the same published clinical guidelines a board examiner would use. The login screen says it plainly: "Students talk to real patients. Before they're real."
Built almost entirely by voice with Claude CodeAccording to Anthropic's official coverage of the Built with Opus 4.7 Claude Code hackathon — where Medkit took first place — Bedirhan built Medkit across four separate Claude Code sessions (voice engine, content generation, 3D game layer, and the core app), following a "talk, don't type" approach and working almost entirely by voice, with Claude Managed Agents doing the heavy lifting.
Real traction, not just a demoThree medical faculties and a pharma company in Istanbul are running pilots, and the project carries 333 GitHub stars. The product is now being operated as a company — this is vibe coding that made it past the demo stage into real clinical training.