[Submission] Avusula Vamshi — MediRAG RAG App
What document did you use and why?
Medical PDFs (blood reports, prescriptions). Patients struggle to understand complex medical terminology, so we built a chatbot to explain documents in plain language.
How does your chunking work?
Paragraph-aware chunking: split by \n\n first, then sentence boundaries, then merge into ~300 word chunks with 50-word overlap.
Which embedding model did you use?
TF-IDF (custom implementation) — lightweight, no model downloads, effective for medical keyword matching.
How to run locally
pip install -r requirements.txt streamlit run app.py Get free Groq API key at console.groq.com/keys
Screenshot
See screenshots/demo.png in the repo
What would you improve with more time?
- Replace TF-IDF with sentence-transformers for semantic search
- Add ChromaDB for persistent vector storage
- Telugu language support using Sarvam AI
- Multi-document comparison across reports
- Voice input for elderly patients