Lighthouse-Free-Medical-Clinic - In Network Doctor Finder

Search in network doctors easily with something as vague as 'Cancer in Buffalo
In network doctors will be displayed sorted by proximity to user
10 doctors displayed per page

AI-Powered Insurance Finder

Inspiration

Finding the right insurance and doctors can be a daunting task. By integrating AI into the process, we aimed to simplify this challenge. Users will be matched with in network doctors and understand complex policies easily just by typing their questions.

What We Learned

AI Integration: We learned how to work with Langchain and Google Gemini API to generate structured outputs from natural language inputs.
FastAPI: This framework helped us build web services efficiently.
NY Gov Health Data API: This API helped us retrieve large volumes of live doctor data.
Cloud Deployment: Using Google Cloud, we gained insight into deploying scalable applications and working with real-time APIs like Google Geocoding and Distance Matrix APIs.
Handling Data: Implementing pagination taught us how to handle large datasets more efficiently for a better user experience.

How We Built It

Frontend: Built with HTML, CSS, and JavaScript to collect user symptoms and display relevant doctors in paginated views.
Backend: FastAPI handles user input and integrates with the Google Gemini API for symptom analysis.
AI-Driven Symptom Analysis: The Langchain template structures user input into prompts for AI to analyze and suggest medical services or specialties.
Distance Calculation: Using the Google Distance Matrix API, the platform calculates the distance between doctors' offices and the user's location and orders results by proximity.
New York Health Data Integration: Leveraged the New York State Department of Health
Deployment: The project was deployed on Google App Engine for better scalability and flexibility.

Challenges We Faced

Structuring AI Outputs: Ensuring AI-generated JSON was consistent required tuning and refining prompts.
Cloud Deployment Issues: Debugging deployment on GCP was tricky, especially with configuring gunicorn and uvicorn.
Geolocation Data: Handling incomplete address data for geolocation and distance calculation was a major challenge.
Pagination: Implementing smooth pagination for large datasets was critical to maintain a clean and responsive UI.

Conclusion

This project combines the power of AI and cloud technologies to simplify the process of finding the right in network doctors, and understand complex policy documents easily. There are multiple scopes for improvement including implementing corrective and adaptive RAG for insurance simplification.