Index / Work / MedTalks — Medical English Platform
MedTalks — Medical English Platform
Platform architecture and development
An English-language training platform for doctors, nurses, pharmacists and dentists, built around AI voice agents that let students practise clinical conversations. Four coordinated applications over one shared data layer.
- Client
- MedTalks
- Industry
- EdTech
- Role
- Platform architecture and development
- Platform
- Web
- Stack
- Flask · Firestore · LiveKit · SocketIO · Twilio
- Dated
- 2026-06-20
- Four independent apps coordinating through one shared datastore
- AI voice agents for clinical conversation practice
- Placement testing with retake cooldown logic
- Live video classes and scheduled meetings
- Credit and minute ledger with regional payment gateway
- Granular per-teacher permissions rather than fixed role tiers
- Course delivery, assignments and progress reporting
① Challenge
Four audiences — students, teachers, admins and the public — with genuinely different needs, plus live video classes, placement testing, credit-based billing and AI conversation practice. Building it as one application would have made every deploy risk the whole platform.
② Solution
Split into four independent applications sharing a single database as the coordination layer. The teacher app holds no data of its own and proxies to the admin backend, sharing a session so a teacher signs in once. The public marketing site uses a deliberately lightweight database client rather than the full SDK, keeping it fast and cheap to serve. Teacher permissions are granular per-resource toggles rather than fixed tiers, so access can be tightened per person without new code.
③ Result
A running platform covering placement testing through to live classes and AI conversation practice, with credit-based billing and regional payment support. Roughly 36,000 lines across the four applications, each deployable independently.