Machine Learning Engineer - ML Systems, Automated Officiating

WORK OPTION: Remote. We are open to candidates able to work in the New York, NY, or Secaucus, NJ, offices.

 

 

Group Summary:

The Automated Officiating team is a new function spanning multiple departments across the NBA, including the Basketball Strategy & Growth and Media Ops & Technology departments. The team’s primary goal is to develop a real-time, multi-modal officiating product – leveraging computer vision and other sensing modalities – used during live NBA games to enhance call accuracy, streamline game flow, and provide decision-making consistency and transparency. This is a new team within the NBA and provides significant opportunities for ownership, accelerated learning, and growth.

 

Position Description:
The NBA is seeking an experienced Machine Learning Engineer to be a key contributor to the Automated Officiating team and be responsible for data infrastructure and ground truth labeling functions. This is a ML Systems and data infra engineering role with scope covering all data infrastructure needed to enable ML iteration at velocity. The ideal candidate will bring experience working on high volume sensor data (e.g., cameras, lidars) and understand the ML data flow from sensing, cloud storage, data compressions, and distributed data processing to produce datasets to train large perception models.  This person will play a critical role in taking our product from 0 to 1, leveraging expertise typically found in autonomous vehicles, robotics, AR/VR, or other real-time ML-driven systems.

 

Major Responsibilities:

• Play a pivotal role in defining the distributed (PB scale) ML data strategy for Automated Officiating.
• Build and maintain data pipelines that handle multi-modal sensor data, including video (high frame rates), sensor feeds, and player and ball tracking data. 
• Optimize the pipeline for storage, compute, and execution velocity. 
• Own the data labeling pipelines and tooling, and work with the Automated Officiating team to define ground truth taxonomies and versioning.
• Collaborate with the Automated Officiating modeling team to integrate perception algorithms into end-to-end officiating solutions. 
• Collaborate with other Media Ops & Technology teams and drive integration of the deployed Automated Officiating outputs into the Replay Center, the broadcast, and other outlets.
• Develop profiling tools to understand performance and data bottlenecks and address.
• Have a strong sense of ownership and be excited to wear many hats.
• Be a guardian of the codebase and push for clean, well-tested and highly extensible code.

 

Qualifications:

• Minimum of 5+ years of experience building production ML data pipelines and/or ground truth labeling and tooling. 
• Experience working with ML data pipelines involving camera, lidar, or other dense sensor input.
• Proficiency in Python and experience with production Machine Learning pipelines: large scale dataset creation, dataset versioning, training frameworks and metrics pipelines.
• Strong grasp of low-latency, high-throughput system design, and distributed computing applied to ML pipelines.
• Familiarity with Cloud providers (AWS, GCP, Azure) and their offerings.
• Excellent problem-solving skills and adaptability in a fast-paced environment. 
• Excellent communication and interpersonal skills.
• Experience building systems to read, synchronize and replay sensing input from a variety of sensors ranging from high-definition cameras to IMUs and IR sensors.

 

Bonus Qualifications:

• Familiarity with video / image compression techniques and experience incorporating decompression into data pipelines.
• Familiar with ML training frameworks (e.g., Pytorch Lightning), and prior experience building ML training and evaluation pipelines.
• Exposure to CUDA, parallel computing, or high-performance programming on GPUs. 
• Background in sports analytics or experience working with sports-specific data.
• Passion for basketball and familiarity with officiating rules.

 

Salary Range: $210,000 - $300,000

 

The NBA does not accept unsolicited resumes from search firms or any other third parties. Any unsolicited resume se