Health Vault

Inspiration

Every time someone we knew got a lab report or hospital bill, they'd call a doctor friend just to understand what it said. Medical documents are deliberately complex — dense jargon, cryptic abbreviations, and billing codes that even trained professionals struggle to parse. We wanted to build a tool that gives every patient the clarity they deserve, without needing a medical degree or a doctor on speed dial.

What it does

HealthVault is an AI-powered medical assistant that puts patients back in control of their own healthcare data. It has four core tools:

• Jargon Buster — Upload any lab report, MRI result, or discharge summary and get a plain-English breakdown of every finding, what it means, and whether you should be concerned.
• Bill Analyzer — Upload a hospital bill and watch HealthVault scan it line-by-line, flagging duplicate charges, inflated fees, and billing code mismatches that could be costing you thousands.
• Prescription Buy — Snap a photo of a handwritten prescription and instantly get a structured medicine list with direct links to buy them online from trusted pharmacies.
• Family Vault — A secure, private dashboard to store, organize, and track health records for your entire family — all protected by Row-Level Security so only you can access your data.

How we built it

• Frontend: Next.js 16 (App Router), React, Tailwind CSS, and smooth Framer Motion animations for a clean, responsive UI that works on any screen size.
• Backend: Supabase handles authentication, PostgreSQL storage, and file uploads — with strict Row-Level Security policies ensuring complete data privacy.
• AI Core: We integrated Google's latest @google/genai SDK using gemini-3.6-flash, which handles multimodal input (images + text) at high speed, letting users upload photos of handwritten prescriptions and scanned documents and get instant structured analysis.
• Testing: Vitest for unit tests across all core logic, and Playwright for end-to-end testing of critical user flows like auth and document uploads.

Challenges we ran into

The biggest technical hurdle was the Gemini SDK migration. Midway through the project, the older model APIs (gemini-1.5-flash, gemini-2.0-flash) were deprecated with 404 NOT_FOUND errors, silently causing our document upload spinner to hang forever. Tracking down why the UI froze with no visible error took significant debugging. We also faced challenges getting Supabase Row-Level Security policies right across shared family health records, ensuring one user could never accidentally read another's data.

Accomplishments that we're proud of

We shipped a fully working multimodal AI pipeline that can read a photo of a handwritten doctor's prescription and extract a clean, structured, buyable medicine list in seconds. The animated UI — especially the homepage button with its spring-physics corner animations — feels polished and genuinely delightful to interact with. We're also proud of the security architecture: a real-world RLS schema that enforces data isolation at the database level, not just the application layer.

What we learned

Integrating cutting-edge AI models in a production-grade app requires building for model deprecation from day one. We learned to always maintain a prioritized fallback list of models (gemini-3.6-flash → gemini-3.5-flash) and to surface backend failures explicitly in the UI rather than letting spinners hang silently. We also deepened our understanding of Supabase's auth flows, specifically the nuances of server-side vs. client-side session handling in Next.js App Router.

What's next for Health Vault

• Voice mode: Let users describe their symptoms verbally and get a structured health summary.
• Doctor comparison tool: Upload multiple prescriptions over time and track how your health metrics trend across visits.
• Insurance claim assistant: Automatically match your hospital bill against your insurance policy and identify what should be covered.
• Offline mode: PWA support so users in low-connectivity areas can still access their stored health records.

Built With

• ai
• computer-vision
• document-analysis
• framer-motion
• gemini-api
• generative-ai
• google-gemini
• healthcare
• javascript
• multimodal-ai
• next.js
• node.js
• playwright
• postgresql
• react
• rest-api
• row-level-security
• supabase
• supabase-auth
• supabase-storage
• tailwind-css
• typescript
• vercel
• vitest