Inspiration
Personal trainers cost $80+ an hour. Most students and everyday gym-goers simply cannot afford that. Bad form meanwhile causes real injuries like knees, backs, shoulders, that sideline people for months. ClaudeGymBuddy was built to put basic coaching in everyone's pocket, for free.
What it does
You record a set, upload the clip, and ClaudeGymBuddy identifies the exercise, counts your reps, scores your form, and flags the exact frames where your technique breaks down. It also highlights injury risks and gives specific corrections but not generic tips.
How we built it
We built the pipeline around Claude's vision capabilities, engineering prompts that extract structured biomechanical data from raw phone footage. The frontend parses and renders the output into a readable coaching report with frame-level breakdowns.
Challenges we ran into
Phone video is messy with inconsistent angles, lighting, and frame rates made reliable analysis difficult. Prompt engineering for consistent, structured output across different exercises took significant iteration.
Accomplishments that we're proud of
Getting accurate form scoring and injury risk detection working on ordinary phone footage with no dedicated hardware or wearable.
What we learned
How to architect an AI pipeline around real-world video input, and how to turn model output into something genuinely useful rather than just impressive.
What's next for ClaudeGymBuddy
Live camera coaching, real-time correction cues between sets, and a coach dashboard for remote trainers to review flagged sessions.

Log in or sign up for Devpost to join the conversation.