Context
Nutrition tracking is crowded, but the core interaction has barely changed: find an item, choose a serving and confirm. If logging a meal feels like admin, users stop doing it.
The problem
- Input friction makes consistent tracking difficult.
- Variable data quality weakens confidence in public food databases.
- Generic onboarding fails to adapt targets to the person.
- Late monetisation work makes paywalls and experiments expensive to retrofit.
The solution
Meal analysis from a photo
The user takes a picture and receives estimated calories, protein, carbohydrates, fat and practical portion guidance within seconds. There is no database search or repetitive confirmation flow.
A product-led onboarding
Sixteen focused screens capture goals, profile, activity, diet and allergies before presenting the subscription at the right point in the journey.
Clean MVVM architecture
Domain, data and presentation layers remain separate. Payment, vision and analytics providers sit behind interfaces, making them replaceable and easy to test.
RevenueCat and remote configuration
Monthly and annual subscriptions run through RevenueCat. Paywall copy, onboarding order and key defaults can be adjusted remotely for controlled experiments.
Architecture
- Expo and React Native with strict TypeScript.
- Supabase Auth and PostgreSQL for accounts and meal data.
- Claude Vision behind a server-side integration and a development mock mode.
- RevenueCat with webhook-based entitlement synchronisation.
- Centralised design tokens ready for light and dark themes.
Next
Apple Health synchronisation, better multi-dish recognition, limited offline support and continued paywall optimisation are the next product steps.