HealthAccess AI – Community Health Agent

Inspiration

In many communities, especially in underserved areas, people struggle to access reliable health information and find nearby clinics. This inspired us to build HealthAccess AI — a multi-step AI agent that bridges the gap between health resources and the people who need them most.

What it does

HealthAccess AI helps users find clinics, check medicine availability, and receive clear health guidance. It uses TiDB Serverless with vector search to store and retrieve health data, and AI to summarize results in simple, local language. The agent can also send information via SMS/WhatsApp for communities without smartphones.

How we built it

• Ingested health documents and clinic/medicine data into TiDB Serverless
• Used vector + text search for smart retrieval
• Integrated AI summarization to simplify answers
• Connected to Google Maps API for directions
• Added Twilio SMS/WhatsApp to reach basic phone users
• Built with Python (Flask) and OpenAI API

Challenges

• Ensuring the system works smoothly on limited devices
• Handling both local languages and English
• Making the workflow multi-step: search → summarize → connect external tools

Accomplishments

• Built a working demo that retrieves and summarizes clinic info
• Designed an inclusive solution that works even on SMS
• Learned how to combine TiDB vector search with agentic AI

What we learned

• How to build multi-step AI agents with TiDB
• The power of combining AI with practical APIs like Maps and Twilio
• How design decisions affect accessibility in low-resource environments

What's next

• Expand to include more health data sources
• Add voice support for people with limited literacy
• Deploy as a full-scale community health platform

Built With

• flask
• google-maps
• openai
• python
• tidb
• twilio