Staff/Sr. Staff Machine Learning Engineer - World Modeling, Simulation

Waymo•Mountain View, CA

11d•$251,000 - $310,000

About The Position

Waymo’s Simulator Team develops realistic simulations for testing, training, and validating the Waymo Driver. The team combines machine learning engineering, software engineering, and research to build simulation platforms using generative and physical world modeling across dynamic agents, road topologies, traffic systems, weather, and vehicle sensors including Camera, LiDAR, and Radar. This role will define and drive the technical roadmap for large-scale foundation world models, connecting multimodal pre-training across 4D video, LiDAR, and Radar with spatiotemporal tokenization and causal post-training or distillation. The goal is to generate interactive, multi-platform, safety-critical driving worlds for closed-loop evaluation and onboard model training. The position reports to a Senior Staff Engineering Manager.

Requirements

PhD or MS in Computer Science, Machine Learning, Robotics, Computer Vision, or a related quantitative field, or equivalent practical research/industry experience.
At least 8 years of industry or relevant research-lab experience developing, deploying, and leading complex ML architectures, or equivalent tenure-track/associate professorship experience.
Demonstrated experience training large-scale generative and physical world models from scratch on multi-node GPU/TPU high-performance compute clusters.
Hands-on experience designing generative-model post-training and student-conversion pipelines, including few-step diffusion/flow distillation, autoregressive exposure-bias mitigation, SFT, and reinforcement learning or reward-model alignment.
Deep technical foundation in 3D computer vision and physical world modeling, including action-conditioned scene dynamics, explicit 3D physical/geometric priors, spatiotemporal tokenization, camera geometry, and generative 4D world representations.
Expert-level Python proficiency and proficiency with modern distributed ML frameworks such as JAX/Flax, XLA/Pallas, PyTorch, Megatron-LM, Pathways, or DeepSpeed for multi-node model, sequence, and data parallelism.

Nice To Haves

Experience at leading video generation labs, foundation model startups, advanced robotics institutes, or high-tier academic research groups focused on physical AI, foundation video models, or world simulation.
Experience generating or synthesizing non-RGB physical sensors, particularly 3D LiDAR point clouds/range images and 4D imaging radar, as well as multi-camera automotive rigs across vehicle platforms and varied weather and lighting conditions.
Experience in physical world modeling and embodied dynamics, including action-conditioned causal world models, differentiable or hybrid physics engines, rigid/articulated body kinematics, vehicle dynamics, interactive multi-agent traffic dynamics, or closed-loop robotics simulators.
Experience with low-level systems optimization, XLA/Pallas TPU kernels or custom CUDA/Triton kernels, block-sparse/Flash/Splash attention over 100k+ token sequences, and high-throughput KV-cached distributed serving.
Working proficiency in modern C++ for high-performance simulation integration.

Responsibilities

Drive the technical vision and long-term roadmap for foundational 4D world modeling, spatiotemporal tokenizers, video generation architectures, and generative sensor simulation across Waymo.
Architect, train, and scale foundation video and physical world models across distributed TPU/GPU clusters; establish training recipes, multimodal data curation strategies, and scaling laws.
Design and implement post-training approaches such as SFT, RLAIF, model merging, and 3D/semantic conditioning to achieve physical realism, stable counterfactual pose divergence, and fine-grained control over safety-critical scene elements.
Develop action-conditioned, 3D-grounded physical world models and multi-view generative pipelines that combine geometric and physical priors with Camera, LiDAR, and Radar synthesis, modeling sensor transport, occlusions, and causal state transitions.
Scale inference and evaluation frameworks into production simulation infrastructure using distillation, quantization, and KV-cached streaming inference for high-throughput closed-loop validation and RL gyms.
Partner with Onboard Perception, Planner, and AI Foundation labs across Waymo and Alphabet to adapt frontier world models into deployment-ready simulation environments for rare-event reproduction, synthetic data training, and closed-loop RL.
Mentor senior engineers and promote high standards for training stability, reproducibility, and system design.