Shakti

Inspiration

Personal trainers give incredible advice — but the moment a session ends, most of it disappears. Trainees forget cues, coaches can't monitor form between sessions, and progress tracking is scattered. We wanted to build the digital extension of every great coach.

What it does

Shakti is an AI-powered fitness coaching platform with two portals — one for coaches, one for trainees. Coaches record sessions via wearable, get transcripts, and use an AI assistant to analyze trainee progress and generate training plans. Trainees upload workout videos for exercise-specific form analysis, view coach feedback, and track improvement over time.

How we built it

We built the frontend as a responsive two-portal web app. The backend handles video processing, audio transcription, and AI integration. Movement analysis runs on MoveNet for joint detection and biomechanical evaluation. Qwen3-8B powers all conversational AI features — session summaries, coaching suggestions, and trainee insights.

Challenges we ran into

Making form analysis truly exercise-aware was hard. Generic posture detection isn't enough — a deadlift and a Romanian deadlift look similar but have very different error patterns. We also had to design a coach–trainee linking system that was secure but seamless to set up.

Accomplishments that we're proud of

Exercise-specific biomechanical feedback that actually tells you why your form is off — not just that it is. And a full dual-portal system where coaches and trainees each have purpose-built tools that work together.

What we learned

AI coaching is only useful when it has context. Generic responses help no one — but when the model knows the exercise, the trainee's history, and past feedback, the insights become genuinely actionable.

What's next for Shakti

Live session analysis, mobile app support, progress benchmarking against training goals, and expanding exercise coverage to include cardio and sport-specific movements.