Inspiration

Modern fitness is broken because data is fragmented. Your workouts, diet, sleep, and environment all exist in isolation, leading to generic and often ineffective health advice.

In India, this problem is even worse due to diverse diets, air pollution, and lack of personalized guidance.

We were inspired to build a system that actually understands a person’s full lifestyle, not just isolated metrics — a system that adapts daily, like a real coach.

What it does

FitIndia is an AI-powered fitness and health tracker that generates a fully personalized daily plan.

It combines:

🥗 Diet (Indian food-based meal plans with macros) 🏋️ Workouts (goal and muscle-group specific routines) 🌫️ AQI-aware recommendations (switches indoor/outdoor workouts) 😴 Sleep, stress, and activity tracking 📊 A 100-point Day Score to guide behavior

The AI dynamically updates recommendations in real-time as the user logs meals, workouts, or lifestyle data.

How we built it

We built FitIndia as a full-stack system:

Frontend: Vanilla JS, HTML, CSS (fast, responsive UI) Backend: FastAPI with async architecture Database: Supabase PostgreSQL for structured health data AI Layer: Groq-powered Llama 3.3 70B for plan generation

Pipeline:

Collect user data (meals, steps, sleep, location) Compute features (TDEE, macros, scores, risk signals) Fetch external data: Nutrition (USDA) Exercises (Wger API) AQI & weather (OpenWeatherMap) Generate structured AI plan (diet + workout) Update in real-time based on user actions Challenges we ran into AI reliability: Preventing hallucinations required structured prompts and controlled outputs Data integration: Combining health, behavioral, and environmental signals into one system India-specific complexity: Handling regional diets and vegetarian preferences Real-time updates: Keeping the system fast while dynamically adapting plans Time constraint: Building a full-stack AI product within hackathon limits Accomplishments that we're proud of Built a working end-to-end AI system, not just a prototype Created a context-aware health engine, not just a tracker Integrated AQI into fitness decisions, which most apps ignore Developed a custom Indian food dataset Achieved real-time adaptive recommendations What we learned Fitness is not just about tracking — it’s about understanding context AI needs structured inputs to be reliable in real-world applications Localization (like India-specific data) is critical for impact Simplicity in UI is as important as backend complexity What's next for FitIndia Integration with wearables (Apple Watch, Fitbit) Doctor and expert-backed recommendations Predictive health insights and risk scoring Expanding Indian food and regional datasets Scaling to a production-ready platform