Computer Vision Research Engineer; Modeling

CV/ML Research Engineer, Automated Officiating

Get AI-powered advice on this job and more exclusive features.

Retrieved from the description.

$/yr - $/yr

WORK OPTION:
Remote

The Automated Officiating team at the NBA is seeking an experienced and senior research engineer with a strong foundation in Computer Vision and Machine Learning, and ideally with Technical Leadership experience to develop advanced computer vision capabilities to automate basketball officiating. This team sits within Basketball Strategy & Growth, and its primary goal is to develop advanced, multi-modal officiating capabilities to enhance call accuracy, streamline game flow, and provide decision‑making consistency and transparency.

This is a small team that works like a startup within the NBA and provides significant opportunities for ownership and accelerated learning and growth. This is a research engineering position; ideal candidates bring considerable expertise in applying state‑of‑the‑art computer vision techniques to reason about scene and player level semantics, player actions and intent, player and ball tracking, and 3D reconstruction and mesh tracking of dynamic objects, with the ultimate goal of building a high‑accuracy system that is able to make live calls for objective violations using cameras and other sensing modalities.

We are looking for candidates that have the skills and aptitude to work on highly complex and ambiguous problems and are excited to contribute to all aspects of a real‑world perception system, from building sensing pipelines to scalable ML data, training, modeling and evaluation pipelines. This role will report to the Engineering Lead and play a critical role in taking the Automated Officiating Product from 0 to 1.

The Basketball Strategy & Growth department is responsible for data collection, analysis and technology pertaining to all on‑court activities. The group, in partnership with Referee Operations, oversees the Game Review Program to help drive improvements in referee performance and rules clarification initiatives. Basketball Strategy & Growth also leads pivotal initiatives focused on innovating and improving the NBA game, such as rules changes, improvements to the competition format and implementation of technologies to improve player health, game integrity and fan engagement.

The Automated Officiating team is a new function within the Basketball Strategy & Growth department. This team is focused on innovating the on‑court product through internally developed and deployed technologies. They spearhead key officiating technology initiatives from concept to launch, leveraging their cross‑discipline expertise in real‑time perception and sensing, computer vision, machine learning, and data analytics. The primary near‑term focus of this team is deploying a system that can automatically detect and determine objective calls (e.g., out‑of‑bounds) in real‑time during live NBA games.

• Designing, implementing, and deploying state‑of‑the‑art tracking, 3D reconstruction and geometry estimation, scene understanding and visual recognition systems.
• Play a role in defining the technical strategy and actively look for problem areas and proactively propose solutions.
• Be a leader and an advocate for good ML design principles and software development practices.
• Stay up to date with the latest literature, technologies, and best practices in computer vision, machine learning and multi‑modal foundation models.
• Provide technical guidance and mentorship to other engineers on the team.
• Make technical contributions across the automated officiating system (e.g. sensor pipelines, ML data pipelines, training, model development and evaluation pipelines).
• Be a guardian of the codebase and push for clean, well‑tested and highly extensible code.

• Masters, or Ph.D. in Computer Science, Computer Engineering, Math or related field (or equivalent professional experience).
• Proven track record to design and implement solutions using modern ML architecture.
• Experience leading projects and driving execution of complex and ambiguous…