Inspiration
In rural India, doctors see 50+ patients daily with unreliable internet and zero time to type queries into drug databases. Medical representatives visit once a month — but drug interactions matter at the moment of prescription. We asked: what if every doctor had a 24/7 AI medical rep in their pocket?
What it does
PharmaBridge is a voice-first AI medical representative that works everywhere:
- Online Mode: Doctors send a WhatsApp voice note asking about drugs, dosages, or interactions. Our cloud AI responds with voice + text in seconds.
- Offline Mode: A lightweight Android app with on-device inference. No internet? No problem. Ask via voice, get instant answers locally.
When connectivity returns, everything syncs — usage logs, updated drug databases, new interactions.
How we'll build it
- WhatsApp Business API for voice note ingestion
- Speech-to-text (Whisper) and text-to-speech pipeline
- RAG architecture with medical knowledge base (drug interactions, dosages, contraindications)
- Quantized on-device LLM (Phi-3 mini) for offline inference
- SQLite + vector DB for local drug storage
- Sync protocol with conflict resolution for seamless online/offline transitions
Challenges we anticipate
- Ensuring medical accuracy with hallucination guardrails
- Optimizing LLM for edge devices with limited RAM
- Handling voice in noisy clinic environments
- Designing sync logic for intermittent connectivity
What makes it unique
No app downloads for WhatsApp mode. No internet needed for offline mode. Voice-native — because doctors don't have time to type. Built for the real world, not demo day.
Built With
- fastapi
- firebase
- langchain
- onnx-runtime
- openai-whisper
- phi-3-mini
- pinecone
- postgresql
- python
- react-native
- sqlite
- whatsapp-business-api
Log in or sign up for Devpost to join the conversation.