Soccer Juggle Tracker — Ball Tracking & Skill Assessment Algorithm | Markana Sport
Computer vision algorithm that counts juggles and predicts player skill level from video.
- Built ball tracking system to count juggles and assess juggling technique
- Integrated sports science research to correlate juggling patterns with player skill level
- Achieved 3x better accuracy predicting 10-attempt averages vs. single-attempt measurement
Stack: Python, computer vision, ball tracking
Jump.AI for Markana Sport (now getrival) — Mobile Vertical Jump Measurement App | Co-Founder & CDO, 2018–2021
Computer vision platform for sports talent evaluation measuring vertical jump from smartphone video.
- Built pose detection and object tracking system achieving <1 inch measurement accuracy
- Secured $250K seed funding
Stack: Python, computer vision, pose estimation, object tracking
Golf Eye — Real-Time Golf Ball Tracking & Swing Analysis App Freelance Project
Built an AI-powered iOS app that tracks golf ball flight from standard phone video and calculates performance metrics—no specialized hardware required
Key Metrics Delivered
- Distance (±8 yards accuracy under 200 yds)
- Ball speed, launch angle, curve (draw/slice in feet)
Results
- Fully offline processing; results in <3 seconds
- Deployed to App Store
Technical Implementation
- Custom YOLO model trained on 8,000+ annotated frames for ball detection (10-15px targets)
- Kalman filter for trajectory prediction and false positive rejection
- Secondary classifier for impact frame detection
- Ground plane estimation using golfer pose + known ball diameter for 3D reconstruction
- Pose estimation for swing path analysis
- Optimized for CoreML: ~40ms inference per frame on iPhone 13
