Project 03 / Independent / AI & sports coaching
Coach Pickle
A pocket pickleball coach that connects conversation, video feedback, and purposeful practice.
From a question to a better practice
Coach Pickle gives pickleball players a place to ask questions, build practice plans, review footage, and track their progress. Different coach personalities make the guidance feel personal, while a drill library connects advice to something players can actually practice.
Connecting the whole experience
I built the mobile product and its supporting services: a Flutter client, a Firebase backend, and a Python computer-vision service. Gemini-backed coaching uses a pickleball knowledge base, and video analysis connects footage with specific observations and follow-up conversation.
The app brings together coaching chat, generated workouts, skill-focused video review, game logging, and progress tracking. Companion watch experiences and home-screen widgets make it easier to record activity beyond the main app.
The engineering behind the coach
The mobile architecture uses Riverpod and typed routing. The backend coordinates coaching requests and video processing, with computer-vision preprocessing using OpenCV, MediaPipe, and YOLO. Local operations tooling supports the knowledge library, screenshots, and practice audio.
Coach Pickle is a separate product from Pickleball Drills, the contract project also included in this portfolio.