AI Fitness App
2025–26 · React Native
An AI fitness coach that plans your training, reads your meals from a photo and talks back by voice — with a Next.js admin panel for the team behind it.
Overview
An AI fitness coaching app built in React Native (Expo) on Supabase. A detailed onboarding feeds GPT-4o, which generates a four-week training plan and then guides every session set by set. Meals are logged from a photo, water, sleep and weight are tracked, and the coach answers by text or by voice over OpenAI's Realtime API. The app is fully bilingual (English and Norwegian), runs subscriptions through RevenueCat, and comes with a Next.js admin panel where the team builds programs with drag-and-drop and manages exercises and users.
Key Features
AI-generated four-week training plans built from onboarding answers — goals, body metrics, equipment, injuries and schedule
Guided workout sessions with warm-up timers, rest periods and set-by-set tracking, plus custom workouts and HIIT
Meal logging from a photo with AI food analysis, alongside water, sleep and weight tracking
AI coach in chat (streaming GPT-4o) and in voice mode over OpenAI's Realtime API, with voice tools that log meals, water, sleep and workouts
Next.js admin panel with a drag-and-drop program builder, exercise library and user management, on the same Supabase backend as the app
Fully bilingual (English and Norwegian), with RevenueCat subscriptions, Sentry crash reporting and EAS builds for iOS and Android
Technical Challenges
Turning AI output into a structured four-week plan the app can actually run, set by set
Building a voice coach that can act — log a meal or a workout mid-conversation — not just talk
Keeping every screen bilingual: Redux stores raw Supabase rows and the UI picks English or Norwegian at render time
Sharing one Supabase schema between the mobile app and the admin panel, with sensitive writes kept server-side