Nutryx

Inspiration

Tracking calories manually is tedious — searching foods, guessing portion sizes, typing everything into a diary. We wanted something closer to how MyFitnessPal solved logging, but removed the typing entirely: point your camera at your plate, and let AI do the counting. Nutryx was built to make nutrition tracking as fast as taking a photo.

What it does utryx lets you photograph any meal and instantly get calories, protein, carbs, and fat — no manual search, no guessing. A short onboarding calculates a personalized daily calorie and macro target based on your goals (lose, maintain, or gain weight), activity level, and desired pace. Every meal is saved to a history you can review anytime, and premium features are unlocked through a subscription paywall powered by RevenueCat.

How we built it

The app is built with React Native and Expo, using Supabase for authentication, database, and file storage. Photo analysis runs through a Supabase Edge Function that calls Claude's vision model to identify food items and estimate nutritional values, returning structured JSON the app renders into an editable result screen. Subscriptions are handled with the RevenueCat SDK, integrated with Google Play Billing. The onboarding flow was inspired by category-leading fitness apps' approach to personalization, adapted and rebuilt from scratch for Nutryx.

Challenges we ran into

Getting the photo analysis pipeline reliable took several iterations — prompt engineering for consistent JSON output, handling images that contain no food, and aligning field names between the AI response and the database schema. We also hit a tricky production bug where environment variables worked locally but weren't bundled into the cloud build, causing the app to fail validation during Google Play review — tracing that down and fixing it for good was a key milestone. Configuring Google Play subscriptions with base plans, offers, and a free trial also required careful testing to get right.

Accomplishments that we're proud of

Shipping a fully working end-to-end flow — photo capture, AI analysis, data persistence, personalized onboarding, and a working subscription paywall — as a solo builder, from scratch, within the Shipaton window. Every piece, from the database schema to the Edge Function to the paywall sequence, was built and debugged hands-on.

What we learned

The gap between "works on my machine" and "works in production" is real — cloud builds don't automatically inherit local environment files, and that single assumption cost real debugging time. We also learned how much of a subscription app's conversion depends on details outside the code: trial length, plan framing, and store configuration all matter as much as the feature itself.

What's next for Nutryx

referral system rewarding both users with extra free days, expanding macro tracking with weekly trends and progress charts, and refining the AI prompt further to improve portion-size accuracy across more food types.

Built With

• api
• claude
• cursor
• gemini
• supabase