Staff Machine Learning Engineer, Perception, Semantics

Within the Perception organization, the Semantics core team is comprised of three specialized subgroups, each tackling a critical aspect of scene understanding. The Vulnerable Road Users (VRUs) subgroup focuses on identifying and interpreting the behavior of pedestrians, cyclists, and other vulnerable road users to enhance safety and rider comfort. The Traffic Control subgroup is dedicated to developing and deploying advanced ML models that interpret traffic lights, signs, and other controls, with a key focus on improving map robustness. The Scene Semantics subgroup addresses the complexities of unstructured environments like construction zones and emergency scenes, driving architectural upgrades and new feature development. Together, these three teams leverage strong end-to-end ML expertise to build robust perception systems essential for safe and efficient autonomous operation.

In this hybrid role, you will report to a Technical Lead Manager.

You will:

• Lead modeling efforts for your respective domain (VRU, Traffic Control or Scene Semantics), encompassing nominal, longtail and fine-grained understanding tasks.
• Leverage onboard and offboard foundation models to streamline your team's model stack.
• Bring up the model stack for rapid adaptation to new tasks and data.
• Establish and maintain a sustainable model stack to facilitate Waymo's multi-platform deployment and scaling goals.

You have:

• 6+ years of experience in Machine Learning, with a strong focus on computer vision and/or deep learning for perception tasks.
• Proven experience leading and mentoring a team of engineers or researchers.
• Deep understanding of state-of-the-art ML techniques for object classification, detection, tracking, pose estimation, and/or action recognition.
• Proficiency in at least one major deep learning framework (e.g., TensorFlow, PyTorch, JAX).

We prefer:

• PhD degree in Computer Science or a similar discipline, or an equivalent amount of deep learning experience.
• Experience with multi-modal perception systems (e.g., combining camera, lidar, radar data).
• Familiarity with foundation models and techniques for model adaptation (e.g., few-shot learning, transfer learning, domain adaptation).
• Experience in optimizing ML models for on-device deployment and real-time performance.
• Background in autonomous driving, robotics, or a related safety-critical domain.
• Publications in top-tier ML/CV conferences (e.g., NeurIPS, ICML, CVPR, ICCV, ECCV).

#LI-Hybrid