FitForge AI
Fitness Coaching
July 2026 – August 2026 · Internship · CNI
Fitness mobile app connecting members and coaches: AI-generated workout programs, nutrition logging by photo or voice, sleep tracking, a performance score and real-time chat, with a web admin console.
By the numbers
0
REST endpoints
0
PostgreSQL tables
0
automated tests
0
exercises with video
Case study
The challenge
Fitness apps keep training, nutrition and coaching apart, and an unconstrained generative AI happily invents exercises or gives risky advice. The goal was to bring everything together in one app, with an AI that is genuinely useful but strictly controlled on the server.
Architecture
Clients
- Flutter app (Riverpod, go_router)
- Next.js admin console
Spring Boot API
- Spring Security + JWT
- REST /api/v1
- WebSocket STOMP
- Layered business services
Data
- PostgreSQL 16
- Flyway migrations
- Media storage
External services
- Gemini
- ExerciseDB
- USDA
- Open Food Facts
Clients
- Flutter app (Riverpod, go_router)
- Next.js admin console
Spring Boot API
- Spring Security + JWT
- REST /api/v1
- WebSocket STOMP
- Layered business services
Data
- PostgreSQL 16
- Flyway migrations
- Media storage
External services
- Gemini
- ExerciseDB
- USDA
- Open Food Facts
Technical choices
01
A controlled AI
Gemini can only pick from a numbered exercise catalogue, so it cannot invent an exercise. The AI coach stays within sport, with an injury safeguard.
02
Protected keys and catalogue
API keys never leave the backend, and the 220-exercise catalogue is frozen in a migration: a fresh database rebuilds without calling the API again.
03
Real time
Coach–member chat runs over WebSocket/STOMP, authenticated with JWT, with photos, voice messages and documents.
04
A tested architecture
Controller → Service → Repository with immutable DTOs, backed by 138 backend tests and 56 Flutter tests.
Key features
- AI-generated workout programs (Gemini) built from a numbered exercise catalog, so no exercise can be invented
- 24/7 AI coach restricted to fitness topics, with an injury safeguard
- Nutrition: USDA and Open Food Facts search, meal photo analysis and voice logging
- 220 exercises with videos, a 2D/3D muscle heatmap and a weekly score out of 100
- Real-time coach–member chat (WebSocket/STOMP) with photos, voice messages and documents
- Next.js admin console: coach verification, reports and AI monitoring
Screenshots (18)
A word about this projectA word about this project
A question, an opinion, an idea?
Write to me right from here. Your message is not published: I am the only one who reads it.
Private: never shown on the site