Canada Basketball

Software Developer & Product Owner · September 2025 — Present

Software Developer & Product Owner – Capstone Project

Sponsored by Canada Basketball

Architecting an automated 3x3 basketball analysis system that leverages state-of-the-art computer vision and machine learning to deliver professional coaching insights from broadcast footage.

Key Responsibilities

• Object Detection & Tracking: Architected an automated 3x3 basketball analysis system leveraging RF-DETR and SAM 2 for high-precision object detection and segmentation, utilizing ByteTrack and custom Vision-Language Models to maintain robust player identity and tracking across broadcast footage.
• Data Processing Pipeline: Developing a sophisticated data processing layer that converts 2D video coordinates to 3D court positions via homography and synchronizes game events with millisecond-precision OCR clock reading to derive professional coaching insights such as shooting efficiency and pace factors.
• Full-Stack Platform: Developed a responsive, high-performance video analysis platform using Next.js, Tailwind CSS, and Docker, featuring a split-screen interface for real-time "Momentum Metrics" visualization and integrating secure Google OAuth authentication.