SearchInsured

Inspiration

Healthcare and insurance are two areas where users are often left frustrated, confused, and overwhelmed. We realized that even simple questions like “What does my plan cover?” or “Which doctors are in-network?” require reading through 70-page PDFs or navigating outdated portals. This frustration inspired us to build SearchInsured — a conversational AI assistant that brings clarity and simplicity to one of the most confusing parts of adult life: understanding your health insurance.

What it does

SearchInsured allows users to ask natural-language questions about their health insurance coverage and find doctors who accept their plan — all through a simple chat interface.

You can ask things like:

“Find cardiologists in Los Angeles.”
“Does my plan cover emergency care?”

The app understands your intent, retrieves accurate results, and presents them in an easy-to-read conversational format. It saves users time, effort, and confusion by turning complex insurance data into direct, human-like responses.

How we built it

We built the frontend using HTML, CSS, and vanilla JavaScript to keep the interface lightweight and responsive.
The backend is powered by FastAPI (Python), which handles user queries dynamically with an AI Agent that has access to two tools:

search_document — allows conversation with insurance PDFs using RAG
search_doctors — performs API calls to the NPI directory to retrieve doctor information

Our /agent endpoint is the single entry point for all requests, and the agent dynamically decides which tool to call based on user intent.

We used FAISS and Sentence Transformers to locally store and retrieve embeddings from insurance plan documents, simulating how the system would process real insurance data.
All components communicate via REST APIs, forming a clean and modular architecture that can scale easily with real data sources.

Challenges we ran into

The biggest challenge was accessing real insurance provider data. Most health insurance APIs are either paid or restricted to partners, which limited our ability to show live plan-specific results. We overcame this by simulating responses using public FHIR structures and focusing on building a fully functional workflow that can plug into real APIs later.

Another challenge was designing an intuitive chat interface that explains complex insurance data clearly — we spent time iterating on formatting, layout, and step explanations to make responses feel natural and trustworthy.

Accomplishments that we're proud of

Built a full-stack, working prototype from scratch within a short time frame.
Created a clean, user-friendly chat interface without relying on any frameworks.
Designed an agent system that not only answers questions but also shows its reasoning steps transparently.
Identified a real, everyday pain point in healthcare and developed a solution that’s practical, empathetic, and impactful.

What we learned

We learned how challenging — and rewarding — it is to simplify something as complex as health insurance.
Technically, we deepened our understanding of FastAPI, vector embeddings, and frontend–backend integration.
More importantly, we learned how to design an AI product that balances technical performance with empathy — making technology truly serve people when they need clarity the most.

What's next for SearchInsured

Our next step is to partner with insurance companies to gain access to their official APIs and provider directories. This will allow SearchInsured to deliver real, plan-specific results at scale.

We also plan to:

Integrate secure user accounts and personalization.
Expand coverage to multiple insurers.
Enhance the reasoning engine for more complex queries.

Ultimately, we want SearchInsured to become the go-to AI companion for understanding and managing health insurance — bringing comfort and clarity to users in times of discomfort.