Inspiration

As someone passionate about fitness and technology, I wanted to solve a common problem many face during workouts: improper form and posture. Poor form not only slows progress but also leads to injuries. I envisioned a mobile-first AI coach that could give real-time feedback and personalized corrections—without needing bulky equipment or expensive trainers. That vision became FitBuddy.

What it does

FitBuddy is an AI-powered fitness coach that helps users master proper form using only their smartphone camera. It uses real-time pose detection to:

Track and score reps based on form quality

Detect posture imbalances (left–right / front–back)

Give instant visual and spoken feedback

Generate adaptive workout plans

Count reps and track progress

Provide habit coaching and injury-risk alerts

All of this happens on-device, keeping the experience seamless and private—no sensors or wearables required.

How I built it

I built FitBuddy using:

React Native with Expo for cross-platform mobile development

TensorFlow.js and Blazepose models for real-time pose detection

React Native Reanimated for smooth UI feedback animations

AsyncStorage for storing user preferences and progress locally

Expo Camera for accessing the front-facing camera

Lucide Icons and Linear Gradients for a modern, vibrant UI

TypeScript for robust typing and maintainability

The architecture is modular, with reusable components for the camera view, heatmaps, rep tracking, analytics, and more.

Challenges I ran into

Fine-tuning pose detection for varying body types and lighting conditions

Preventing frame drops during high-load computation (real-time detection)

Designing an interface that conveys complex feedback in a clear, motivating way

Managing app state and camera permissions smoothly on both iOS and Android

Syncing motion logic with animations while keeping performance high

Accomplishments that I'm proud of

Creating a completely sensor-free fitness tracking experience

Developing real-time rep quality scoring and imbalance detection

Designing a sleek, motivational UI that's beginner-friendly yet powerful

Achieving on-device analysis for privacy and responsiveness

Building the full project solo from scratch

What I learned

Advanced animation techniques using Reanimated and shared values

Managing local data with AsyncStorage in a clean and reactive way

Performance optimization for on-device ML models in mobile apps

UI/UX best practices for fitness applications

Adapting open-source pose models to custom use cases

What's next for FitBuddy: AI Fitness Coach & Form Tracker

Cloud sync and personalized insights via a backend (e.g., Supabase or Firebase)

More exercise types and custom workout creation

A voice assistant that guides workouts hands-free

Leaderboards, challenges, and community engagement features

Offline capabilities for workouts without an internet connection

Built With

Share this project:

Updates