Inspiration
Accessing reliable medical information is overwhelming—patients struggle to interpret research, while doctors spend hours scanning through papers. We wanted to bridge the gap by combining vector search, AI summarization, and real-time recommendations into one assistant that makes healthcare knowledge more accessible.
What it does
HealthBot ingests medical PDFs, PubMed articles, or patient health logs into TiDB Serverless. It then uses vector search to find the most relevant studies based on symptoms, summarizes them into easy-to-understand insights using Hugging Face models, and recommends nearby doctors or pharmacies via free APIs.
How we built it
Database: TiDB Serverless to store documents, embeddings, and user data.
Backend: FastAPI with routes for ingestion, search, summarization, and external API integration.
AI Models: Hugging Face models (MiniLM for embeddings, BART/Flan-T5 for summarization).
Frontend: ReactJS + TailwindCSS for a simple, clean UI with login, upload, and search dashboards.
APIs: Free healthcare APIs (OpenFDA, BetterDoctor free tier) for recommendations.
Challenges we ran into
Managing embeddings efficiently in TiDB’s vector store.
Choosing free, reliable APIs for doctor/pharmacy info.
Optimizing Hugging Face inference speed.
Ensuring summaries were accurate yet understandable for non-experts.
Accomplishments that we're proud of
Built a multi-step AI workflow fully integrated with TiDB Serverless.
Translated complex research into patient-friendly insights.
Delivered an end-to-end app with database, backend, AI, and frontend working seamlessly.
Created a solution that has social impact in healthcare accessibility.
What we learned
How to leverage TiDB Serverless for vector search in real-world use cases.
Efficiently chaining AI models with FastAPI.
The importance of designing simple, trustworthy user flows for sensitive domains like healthcare.
How external APIs can enrich AI workflows with actionable next steps.
What's next for AI-Powered Personalized Medical Research Assistant
Add multi-language support for global accessibility.
Fine-tune medical summarization models for higher accuracy.
Build a mobile app version for patients on the go.
Integrate with wearable/sensor data for real-time personalized insights.
Partner with open-source medical datasets to expand research coverage.
Log in or sign up for Devpost to join the conversation.