Fit UP

Inspiration

Every new year, millions of people set fitness resolutions, but motivation often fades after just a few weeks. Most fitness apps actually slow you down, forcing you through endless forms, taps, and logging before you can even start moving. We wanted to flip that experience.

Our inspiration was to build a coach that feels natural and frictionless: minimizing input, guiding you in real time, and automatically handling the boring parts. Fit AI gives instant feedback on whether your reps are correct, how much intent and energy you’re putting in, and then auto-summarizes the session so you don’t waste time logging it. The result is a system that actually helps people stick with their fitness journey rather than drop off after a month.

What it does

Fit AI is a hands-free, AI-powered, real-time fitness coaching system. It acts like a personal trainer in your pocket by:
• Generating personalized workout and meal plans instantly.
• Tracking motion and form during workouts with computer vision.
• Giving instant feedback and rep counting in real time.
• Providing smart analytics, progress tracking, and post-session summaries.
• Offering motivational boosts, like tailored music and interactive coaching.

How we built it

• AI Fusion: Combined computer vision (pose estimation with MediaPipe & OpenCV), predictive models (calorie expenditure, random forests), and generative AI (Groq’s LLaMA-3).
• Frontend: Built with Vite + React, visualized insights using Chart.js.
• Backend: FastAPI served as the backbone for AI orchestration.
• Data Handling: Used structured JSON, OpenWeatherMap for context-aware recommendations.
Core Features:
• Voice-first onboarding using natural speech.
• CV scripts for seamless real-time session execution.
• Physiology-aware calorie tracking calibrated with ML.
• Adaptive meal prep scheduling aligned to predicted calorie burn.

Challenges we ran into

• Accurate rep segmentation across body types and camera angles.
• Low-latency camera handling on Windows (threaded grabbers, buffer control).
• Designing human-readable feedback from raw kinematics (angles/velocities).

Accomplishments that we're proud of

• Created a fully working AI-powered coaching pipeline from onboarding to post-session summaries.
• Achieved real-time form feedback and rep counting with on-device computer vision.
• Developed adaptive plans that integrate both workouts and nutrition dynamically.

What we learned

• How to fuse different AI paradigms (CV, predictive, generative) into a single user-facing product.
• Technical lessons in optimizing real-time AI pipelines, especially for video-based pose estimation.

What's next for Fit AI

Near-Term Expansion:
• Add new exercises (e.g., deadlifts, pull-ups).
• Integrate wearables for live heart rate data.
• Launch white-label platforms for gyms (B2B).
• Enable competitive and cooperative online workouts.
Data Ecosystem & Growth:
• Launch FIT-AI Pro with advanced analytics and nutrition.
• Partner with fitness brands for affiliate gear recommendations.
Advanced AI Features:
• Predictive injury risk analysis from form breakdown.
• Athlete talent ID platform using anonymized benchmarks.