MedAI Hospital

Inspiration

In Bangladesh, many people find it hard to understand their lab reports, find the right medicine information, or book a doctor quickly. We wanted a single place where patients get simple, Bangla and English health guidance, while a real doctor always stays in charge of anything medical.

What it does

MedAI Hospital has three portals:

Patient: request appointments, chat with doctors and admin, upload reports for AI analysis, search a Bangladesh medicine catalogue, and ask the MedAI Assistant health questions.
Doctor: manage appointments and open slots, review AI-analysed reports, write prescriptions with AI-suggested medicines and tests, and approve a signed PDF.
Admin: verify doctors, manage schedules, users, medicines, diseases and banners, review conversations through an audit log, and create temporary access links.

Reports follow one flow: the patient uploads, AI writes a simple-language summary, the doctor reviews and approves, and the patient receives a PDF with a QR code. AI output is only a suggestion for the doctor. It is not a prescription.

How we built it

A Node.js and Express backend with MongoDB Atlas (GridFS for files), a TypeScript frontend, AI models through OpenRouter, and deployment on Render. Calls and chat use WebRTC, and nearby care uses OpenStreetMap data. An Android APK is also available.

Challenges we ran into

Fixing file storage errors so uploaded PDFs could be analysed reliably.
Separating prescription reading from lab report analysis.
Keeping patient passwords and admin access secure.
Making the AI output clear, with a visible approval status from draft to doctor-approved.

What we learned

In healthcare, the AI must support the doctor, never replace the doctor. Clear approval steps, audit logs and honest disclaimers matter as much as the model itself.