Computer Vision Engineer

Company Description HomeTeam Network (HTN) delivers streaming technology for professional, collegiate, Olympic, youth, and high school sports organizations. The company produces thousands of live and on-demand events every year across 42 different sports, serving audiences in Canada, the United States, and Europe. HTN focuses on reliable, high-quality sports broadcasting that scales from elite competitions to grassroots events. Team members collaborate with diverse sports partners and work with cutting-edge media technology. Joining HTN offers opportunities to influence how sports content is captured, processed, and experienced globally.
Employment type: Full-time, regular Employee
Compensation: 85k - 150k plus Equity participation
Role Description You'll build and run the computer vision behind those products: detection, tracking, automated camera framing and event recognition in live sports video. You'll own how it behaves in production, not just how it scores offline. The footage is real-world: variable arena lighting, glare, netting and glass, small fast objects, imperfect camera placement and limited venue bandwidth. Models run in real time on GPUs in the cloud and at the edge, and people watch the output live.
What you'll do
Design, train and ship detection and tracking models for players, ball or puck, and game events across several sports
Build automated camera framing that follows play at broadcast quality
Build video-based event detection for highlight generation, and evaluate it against our existing detection methods
Optimize real-time inference for GPU and edge hardware under latency, compute and bandwidth limits
Own the data loop: capture from our own game footage, annotation, dataset versioning, and targeted collection when a failure mode shows up
Define evaluation that reflects what a viewer notices (a missed play, bad framing, the wrong moment in a highlight) alongside standard metrics
Monitor models in production across venues, sports and lighting conditions, and fix what drifts
Support camera calibration and venue geometry
Build tooling that lets operations staff diagnose and clear issues during a live event
Work with streaming, platform, product and customer-facing teams on what partners need
Required
Experience shipping computer vision models to production and keeping them running
Strong Python and hands-on PyTorch or TensorFlow
Solid grounding in object detection and multi-object tracking
Experience with event or action recognition in video, including finding when an event starts and ends
Real-time video pipeline experience with OpenCV, FFmpeg or GStreamer
Comfort with messy real-world footage: occlusion, motion blur, variable lighting, compression artifacts
Willingness to own a problem end to end on a small team, from data through deployment
Clear written English and experience working with distributed teammates
Nice to have
Sports video experience: broadcast automation, player or ball tracking, sports analytics
GPU and edge deployment: TensorRT, ONNX, NVIDIA Jetson, quantization
Camera calibration, homography, multi-camera geometry or PTZ control
Audio event detection or audio-visual fusion
Annotation tooling, active learning or semi-supervised learning
MLOps: experiment tracking, dataset and model versioning
Cloud deployment on AWS or GCP
Hands-on use of AI coding tools
Familiarity with hockey, baseball, football, soccer or basketball
Working Spanish

Work arrangement
Yes