Fitness Tracker AI

Inspiration

Fitness apps today often feel overwhelming or impersonal. Most of them either give the same generic routines to everyone or require deep fitness knowledge to use properly. As students and beginners ourselves, we wanted something that could understand our situation instantly. Our goals, our schedule, and our limits, without complicated setups or trainers. That idea led us to Fitness Tracker AI: a simple, intelligent fitness companion that creates personalized workout plans in seconds using AI.

What it does

Fitness Tracker AI generates a fully personalized workout routine based on user input such as fitness goals, experience level, available days per week, and equipment. The app: Collects basic fitness details through a clean UI Uses AI to generate a structured workout plan Displays workouts day-by-day with exercises and rest periods Tracks progress as users complete exercises Adapts over time using feedback like “Too Easy” or "Too Hard" or completion streaks All of this happens in real time, without requiring the user to understand fitness theory or planning.

How we built it

We built Fitness Tracker AI using a client–server architecture: Frontend: Vanilla HTML, CSS, and JavaScript Single Page Application (SPA) Dark fitness-themed UI with animations and responsive design Handles user input, UI state, and workout rendering Backend: FastAPI (Python) for speed and clarity Pydantic models for strict data validation Google Gemini API (gemini-2.0-flash) for AI-generated workout plans Custom prompt engineering to force structured JSON output Local persistence using in-memory storage + JSON files The frontend communicates with the backend via REST APIs, keeping the system modular and scalable.

Challenges we ran into

Getting consistent AI output: Large language models can be unpredictable. We had to carefully design prompts and validation logic to ensure the AI always returned usable, structured workout data. State management in vanilla JavaScript: Without frameworks, managing user state, progress tracking, and UI updates required careful planning. Time constraints: Building a full AI-powered product — frontend, backend, storage, and deployment — within hackathon time pushed us to prioritize smart architecture and fast iteration.

Accomplishments that we’re proud of

Building a fully functional AI fitness app, not just a prototype Successfully integrating Gemini AI with strict JSON validation Designing a clean, responsive, and animated UI using only vanilla technologies Creating a system that feels useful, realistic, and scalable Completing the project end-to-end: idea → product → demo

What we learned

How to design reliable AI systems using prompt engineering and validation How to structure a real-world FastAPI backend How frontend and backend systems communicate in production apps The importance of UX even in technical projects How to work efficiently as a team under pressure Most importantly, we learned how to turn an idea into a working product.

What’s next for Fitness Tracker AI

Future improvements we’re excited about: A more intuitive UI Adaptive workouts that evolve based on long-term progress Nutrition and recovery recommendations User accounts with cloud sync Integration with wearables and fitness sensors Smarter feedback loops using AI analysis Fitness Tracker AI has the potential to grow into a true AI fitness companion, not just a workout generator.

Built With

• cors-middleware
• css3
• gemini-api
• html5
• http-rest-api
• javascript
• pydantic-data-validation
• python
• uuid
• uvicorn