AI Calisthenics Coach — Hacktoberfest Weekend Challenge

Devkrishna Sahu — Mon, 05 Oct 2026 07:00:10 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built the AI Calisthenics Coach, an open-source, AI-powered personal trainer featuring live 3D posture tracking, real-time rep counting, and personalized audio debriefs.

I built this for my budget-conscious college friends who are dedicated to calisthenics and bodyweight training but cannot afford expensive gym memberships or human coaches. When working out in dorms or at home, it is incredibly difficult to know if your form is correct. This app solves that by turning a standard laptop webcam into a strict, intelligent coach that evaluates joint angles (like push-up depth), counts valid reps, and even enforces an anti-cheat "Foul Mechanic" if they try to skip the assigned exercise.

Demo

Live App: [PASTE YOUR LIVE LINK HERE - IF NOT READY, DELETE THIS ENTIRE DEMO SECTION]

Code

Devkrishnasahu07 / calisthenics-coach

How I Built It

The frontend and gamified state management are built in Python using Streamlit, styled with custom CSS for a premium dark-mode aesthetic.

For the core engine:

Computer Vision: I utilized open-source OpenCV and Google MediaPipe to perform real-time 3D skeletal landmark tracking. It calculates the angles between the shoulder, elbow, and wrist to track movement phases.
Local AI Brain: I used the open-weights Gemma 2 (2B) model running entirely locally via Ollama. The CV module passes rep counts and foul data to Gemma 2, which then generates a personalized workout debrief.
Memory & Voice: Integrated Backboard to give the AI long-term memory of a user's progress, and ElevenLabs to synthesize the coach's voice.

Why Does Open Innovation Matter?

Open innovation was the only way this project could exist for its target audience. By utilizing local, open-weight models like Gemma 2 via Ollama, I was able to build a tool that my friends can use completely free of charge, without racking up massive API bills for every workout chat.

Processing the webcam feed using open-source tools like MediaPipe ensures that a user's personal fitness and video data never leaves their local machine, solving massive privacy concerns associated with sending video to closed-source cloud APIs.

Prize Categories

Best Use of Local AI / Open-Weight Models
Best Use of Gemma

DEV Community: Devkrishna Sahu