Sports Computer Vision Engineer

This is a fully remote position, open to applicants in United States.

📋 Description

• Develop and train computer vision models specifically for sports video, which includes tasks such as player/ball detection, multi-object tracking, pose/keypoints analysis, event/action recognition, and identity association (re-ID).

• Take ownership of the experimentation process from formulating hypotheses and conducting ablations to performing error analysis and implementing measurable improvements.

• Create and uphold evaluation metrics that are suitable for tasks, along with maintaining dataset slices and failure taxonomies.

• Enhance data efficiency by employing augmentations, sampling strategies, managing label noise, and utilizing weak/self-supervision techniques.

• Design prototypes and iterate on transformer-based detection/tracking models, temporal models, and multi-task architectures.

• Collaborate on the design of datasets and labeling, encompassing formats, schemas, tooling, and version control.

• Assist in the productionization of models through effective packaging, batch/stream inference patterns, and evaluating throughput/latency trade-offs, as well as ensuring robustness checks.

• Implement quality gates to guarantee reproducibility, automated evaluation, and regression detection.

• Work in partnership with data and platform team members to ensure the reliable deployment of computer vision solutions.

⛳️ Requirements

• Demonstrated applied experience in computer vision with practical model development, rather than merely executing existing repositories.

• Proficient in PyTorch, with skills in training loops, debugging, vision data pipelines, and basic distributed data parallelism (DDP).

• Familiarity with fundamental concepts in video computer vision, including occlusion, identity switches, temporal consistency, calibration, and domain shifts.

• Strong Python programming skills coupled with a focus on achieving measurable results.

• Experience in sports video computer vision or related fields is an advantage.

• Familiarity with FFmpeg, WebDataset/shards, or streaming/batching processes to GPUs is a plus.

• Background in MLOps or production environments involving model packaging, continuous integration for training/evaluation, serving via Triton/TorchServe, and monitoring is advantageous.

🏝️ Benefits

• Competitive Salary and Bonus Plan

• Comprehensive health insurance plan

• Retirement savings plan (401k) with company match

• Remote working environment

• A flexible, unlimited time off policy

• Generous paid holiday schedule - 13 in total, including the Monday after the Super Bowl

• Annual performance bonus

• Benefits and/or other applicable incentive compensation plans