Inspiration
Many wheelchair users go years without being weighed or examined on a table, because their clinic has no accessible scale or exam table. Deaf patients arrive to find no interpreter booked. Autistic patients walk into loud, bright waiting rooms with no warning. In 2026, new federal rules (HHS Section 504 and DOJ ADA Title II) started requiring most providers to have accessible exam tables and scales. Patients still have no way to see which clinics do. We built CareFit so disabled patients can find a doctor who can examine them before they show up.
What it does
CareFit is a phone-first web app that matches a patient's access needs to clinics that meet them. Profile: you choose your access needs across mobility, communication, sensory/cognitive, and other. Each one is marked required or nice to have. You also set your location (GPS or ZIP code) and a search radius. Everything stays on your phone. Search: filter by type of care, distance, and needs, or ask the AI assistant in plain language, e.g. "OB-GYN within 15 miles who can weigh me seated." Match: each clinic gets a fit score with reasons in plain text, e.g. has: wheelchair scale · missing: ASL interpreter · unknown: patient lift. Results appear as a list or on a map. Act: open a clinic's details, then create an accommodation request letter that cites Section 504 and the ADA. You can copy, share, email, or print it. The demo uses about 20 fictional Oakland clinics, labeled Demo data. Public data on clinic equipment doesn't exist, and that gap is the problem CareFit addresses.
How we built it
We spent the first phase defining the problem, the users, and the features. Then we designed mockups for every screen and checked them for color contrast. Next came a step-by-step build plan. Claude Opus 5.5 helped us plan, and Claude Sonnet 5.5 agents wrote the code in parallel, one per group of screens. We reviewed and tested each part. The app uses Vite, React, and TypeScript, with Leaflet and OpenStreetMap for the map. A small Node server connects the assistant to a local Ollama model, so no messages go to an AI company. We target WCAG 2.2 AA accessibility and test with Playwright and axe.
Challenges we ran into
We spent more time than planned defining the problem and deciding what to build during the hackathon. We had to build with clearly labeled demo data rather than make claims about real clinics and it took us a bit time and planning to confirm our mockup designs before we built the official submission.
Accomplishments that we're proud of
We built a working platform for a real problem: disabled patients often can't tell whether a clinic has the equipment or support they need until they arrive. CareFit matches each user's access needs, such as an accessible exam table, an ASL interpreter or a quiet room, against a library of clinics, shows the matches on a map and drafts an accommodation request letter. We shipped it as a full-stack, phone-first web app with an AI assistant that runs on a local model, so users' data isn't sent to an outside AI company. We also used Claude to help shape the architecture and to plan how users would find and use these tools.
What we learned
We learned to stay calm and focused under a tight deadline while keeping the people we were building for at the center. Our process worked best in this order:
Brainstorm the idea as a team. Research the problem and find real examples, such as wheelchair users who go years without being weighed because their clinic has no accessible scale. Review our codebase, then hold focused sessions with Claude on specific questions and code snippets we weren't sure about. Asking narrow, specific questions gave us much better results than asking broad ones.
What's next for CareFit
More data sources: connect APIs and public provider directories to make clinic information more complete and accurate. Community reports: let patients report when a clinic's accessibility doesn't match its listing. Real clinic data: replace our demo clinics with verified listings, starting in one city.
Log in or sign up for Devpost to join the conversation.