You can upload a report and medpal converts that it into vectors.

🩺 MedPal

MedPal is your personal medical buddy. Don't know where to save so many medical reports or how to understand them when needed? Don’t worry — we got you.

MedPal simplifies complex medical jargon into easy-to-understand summaries. It checks whether your medical values fall within normal ranges and suggests actionable next steps, helping you seek timely assistance from medical professionals. It also organizes your reports and lets you query them using Retrieval-Augmented Generation (RAG).

💡 Inspiration

We noticed how overwhelming medical reports can be — full of jargon, numbers, and reference ranges that aren’t easy to interpret. People often have no idea what to do after receiving lab results. We wanted to create something that not only stores reports but also makes sense of them, offering guidance in plain language.

⚙️ What It Does

MedPal lets users:

📄 Upload medical PDF reports
🧠 Extract and analyze data from those reports
🔍 Check if medical values fall within normal ranges
💬 Ask questions like “Is my blood sugar level normal?”
🗒️ Get actionable suggestions like “Consult a cardiologist”
📚 Store and view all past medical reports

All of this is powered by a Retrieval-Augmented Generation (RAG) system integrated with embeddings and an LLM.

🛠️ Tech Stack

Frontend

Next.js – UI and frontend interactions

Backend

FastAPI – For file handling and query endpoints
Cohere – For generating embeddings
Qdrant – As a vector database for chunked PDF content
Groq – To run the LLM and generate responses

🚀 How It Works

🔐 Sign in with Google
📤 Upload a medical PDF file
💬 Ask questions about your report using the built-in RAG
🗂️ View your report history and query past uploads

🧗 Challenges We Ran Into

The biggest challenge was developing the RAG pipeline — this was our first time building something like this. We started with only a basic idea of what a vector is. Concepts like chunking, embeddings, and semantic search were all new to us, and we learned as we went.

We also faced:

Issues with deprecated packages (especially in LangChain)
Problems related to environment setup
Difficulty getting vector-based filters to work correctly with metadata

🏆 Accomplishments That We're Proud Of

We’re proud that we:

Built a fully functional RAG system end-to-end
Successfully stored and retrieved document chunks from Qdrant
Created a clean, usable frontend for uploading and querying reports
Learned a LOT in the process!

📚 What We Learned

Environments should be isolated and managed properly
Deprecation warnings are not to be ignored — they matter
RAG is powerful, but also a bit complex
Metadata filtering in vector databases like Qdrant is crucial for multi-user support

🔮 What’s Next for MedPal

Chat Memory – Add session memory so chats feel more natural and context-aware
Personalized Health Trends – Track changes in values across time
Report Categorization – Group reports by type or department