Machine Learning Engineer - Scene Understanding

Zoox•Boston, MA

17h•$179,000 - $245,000

About The Position

The Perception Scene Understanding team at Zoox builds the high-performance reasoning engines that allow our autonomous vehicles to navigate complex urban environments. We translate sensor data and detected objects into deep semantic understanding, ensuring our robots make human-level decisions in real-time. We are seeking engineers passionate about the intersection of robotics and cutting-edge AI. In this role, you will develop next-generation algorithms and deploy production-grade software directly to our fleet. You will tackle the inherent unpredictability of urban drivingâincluding road closures, occlusions, dynamic traffic controls, and abnormal traffic patternsâensuring our vehicles remain resilient in the face of unexpected hazards. As an engineer on the Scene Understanding team, you will develop and deploy advanced machine learning models that interpret the robotâs surroundings to identify hazards and enforce driving restrictions. Your work will be pivotal in resolving unstructured scenarios, ensuring safe navigation through complex or restricted zones. You will optimize models for real-time inference on our vehicle hardware and collaborate closely with teamsâincluding Localization, Perception, Prediction, and Planningâto deliver a seamless, safe, and reliable experience for our riders.

Requirements

BS, MS, or PhD in Computer Science, Robotics, or a related field with 2+ years of industry experience training and deploying ML models.
Experience with BEV (Birdâs Eye View) models, segmentation, or 3D representations.
Hands-on experience with the full lifecycle: dataset construction, training frameworks, and building metrics pipelines to evaluate performance.
A strong mathematical foundation in linear algebra, calculus, and probability.
Fluency in C++ and Python. Experience with PyTorch, TensorFlow, JAX.

Nice To Haves

Track record of designing novel neural network architectures or working with probabilistic outputs.
Experience working on autonomous systems or integrating ML models into a robotics stack.
Publications in top-tier conferences like CVPR, ICCV, ECCV, RSS, or ICRA.

Responsibilities

Develop machine learning algorithms for Birdâs Eye View scene analysis and real-time hazard detection.
Design and implement scene understanding solutions specifically for Zoox's unique bidirectional vehicle.
Lead the end-to-end data strategyâincluding mining, auto-labeling, and dataset constructionâto keep our ML models improving.
Implement on-vehicle software for model inference and result post-processing that runs reliably in real-time.
Use our large-scale data pipelines and infra to research, prototype, and deploy changes that directly improve how the vehicle handles complex road scenarios.
Team up with Perception, Prediction, and Planning. teams to integrate your signals into the final driving solution.