Inspiration
Medical students face two overwhelming daily challenges — digesting massive textbooks and interpreting complex lab reports. We wanted to build something that doesn't just show information, but actually teaches. That's what inspired LearnMed AI.
How We Built It
We built LearnMed AI using:
- Google Gemini 2.0 Flash as the AI brain for multimodal understanding
- FastAPI (Python) for the backend API
- HTML + CSS + Vanilla JS for a clean, fast frontend
- PyPDF2 for PDF text extraction
- Pillow for medical image processing
The app has three core modules:
- Study Assistant — Upload any medical PDF → get summary, key concepts, and auto-generated quizzes
- Medical Report Analyzer — Upload lab reports → detect abnormalities with 🔴 Critical / 🟡 Borderline / 🟢 Normal color coding + educational insights
- Smart Chat — Unified AI chat connecting both study material and medical reports
What We Learned
- How to use Gemini's multimodal capabilities for both text and image analysis
- Building async FastAPI endpoints for file uploads
- Structuring AI prompts to return consistent JSON output
- How to make complex medical data understandable for students
Challenges We Faced
- Getting Gemini to return consistent JSON every time required careful prompt engineering
- Handling both PDF and image medical reports in a unified pipeline was tricky
- Designing the abnormality detection system with proper color-coded severity levels
- Making the UI fast and responsive without any heavy frameworks
What's next for LearnMed AI
- Real-time voice interaction — Ask questions about your notes using voice, powered by Gemini Live API
- Multi-language support — Explain medical reports in Hindi, Telugu, Tamil for regional medical students
- Collaborative study rooms — Students study the same PDF together with shared AI chat
- Progress tracking — Track quiz scores and weak topics over time with personalized recommendations
- Hospital integration — Help junior doctors quickly understand patient reports during rounds
Built With
- css3
- fastapi
- gemini-api
- google-gemini-2.0-flash
- html5
- javascript
- pillow
- pypdf2
- python
- uvicorn
