TrainerAI is a full-stack fitness training platform designed for personal trainers and serious lifters. It combines structured workout management, client tracking, analytics, and AI-assisted workout generation into a clean, production-ready web application.
🔗 Live App: https://trainer-ai-rouge.vercel.app/
- Secure authentication with Firebase Auth
- Email/password signup & login
- Forgot password flow
- Automatic trainer account provisioning on first login
- Full trainer isolation (no cross-user data leakage)
- Create, edit, and delete clients
- Each client is scoped to the logged-in trainer
- Clean empty states for new users
- Create workouts manually or via AI
- Modular block-based workout editor
- Add/remove exercises, sets, reps, weight, tempo, rest
- Reorder blocks and sets
- Workout status tracking (draft, planned, in progress, completed)
- Fully persisted and editable workouts
- Generate structured workouts using Google Gemini
- Uses existing exercise database (global + user exercises)
- Graceful handling of partial/invalid AI outputs
- Generated workouts are editable before saving
- Clean dashboard with:
- Total workouts
- Total exercises
- Total clients
- Recent workouts
- Weekly status breakdown
- Most-used exercises
- Smart empty-state UX for new users
- ~44 globally seeded exercises available to all users
- Trainers can create their own custom exercises
- Global exercises are read-only, trainer exercises are editable
- Safe validation to prevent accidental global edits
- Centralized API error parsing
- Actionable, user-friendly error messages
- No generic “Network Error” screens
- Safe behavior when backend is temporarily unavailable
- Clean loading & empty states
- Next.js (App Router)
- TypeScript
- Tailwind CSS
- Axios
- Firebase Auth (Client SDK)
- FastAPI
- SQLAlchemy
- Pydantic v2
- Firebase Admin SDK
- Google Gemini API
- PostgreSQL (Render)
- Trainer-scoped relational data model
- Global + trainer-specific exercise support
- Frontend: Vercel
- Backend: Render
- Auth: Firebase
- AI: Gemini
- All API routes protected with Firebase ID tokens
- Trainer resolved server-side on every request
- Strict trainer scoping at the database layer
- No client-side trust for user identity
- Backend cold starts handled gracefully (Render free tier)
- CORS configured for Vercel domains
- Environment-based configuration for all secrets
- Production-ready build with strict TypeScript checks
- Multi-account isolation verified
- CRUD flows tested for:
- Clients
- Workouts
- Exercises
- AI generation tested end-to-end
- Dashboard analytics validated
- Auth flows tested (signup, login, logout, reset)
- Program templates
- Workout scheduling/calendar
- Client-side progress tracking
- Export/share workouts
- Mobile-first polish
- Role-based permissions (coach vs athlete)
TrainerAI is built to demonstrate:
- Production-grade full-stack architecture
- Secure auth & multi-tenant data isolation
- Real-world AI integration (not a demo toy)
- Clean UX with attention to edge cases
- Deployment, debugging, and hardening experience
Built by Varun Mantha
Computer Engineering @ Rutgers University
Focus areas: Full-stack engineering, AI systems, production software
If you find bugs, edge cases, or have ideas — feel free to open an issue or reach out.