Machine Learning Engineer, Simulation Realism (World Model)

The Simulator Team at Waymo builds state-of-the-art simulations of realistic environments for testing and training the Waymo Driver. Our team is a diverse and collaborative group of software engineers, machine learning (ML) engineers, and data scientists. We develop industry-leading simulation solutions using advanced ML algorithms that measure and enhance the performance of the Waymo Driver. By applying machine learning, we model the real world, including realistic agents (vehicles, pedestrians, cyclists, motorcyclists), roads, traffic control systems, and weather conditions.

To increase the fidelity and steerability of our simulations, we employ the latest ML technologies such as large foundational World Model trained on our datasets, world understanding and reasoning capabilities for tail cases, and reinforcement learning for long term credit assignment. We also invest in capable infrastructure that allows us to quickly set up and roll out multiple counterfactual scenarios to rigorously test our autonomous driving systems.

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

You will:

• Drive innovation in ultra-realistic world simulation for autonomous vehicles by developing and scaling foundation models for embodied agents.
• Design and train multimodal world models and generative architectures using diverse sensory data (such as RGB images, video, and text).
• Drive closed-loop training and reinforcement learning pipelines to advance simulation realism, policy optimization, and decision-making for autonomous driving.
• Work with Waymo's foundational AI research team to transfer research into scalable, production-ready solutions.

You have:

• 5+ years of experience with ML software engineering
• 5+ years of experience with Python
• 5+ years of ML software experience with machine learning frameworks including PyTorch, TensorFlow, or JAX/Flax
• Demonstrated experience building, training, or scaling world models or multi-modal models on large-scale datasets (e.g., video, images, text)
• Hands-on experience with closed-loop training, policy learning, or reinforcement learning frameworks

We prefer:

• Experience in robotics, control theory, or embodied AI (strongly preferred, but not required)
• Experience working with, creating, and/or developing ML models for autonomous driving systems
• Strong C++ in a production setting, with experience in integrating ML models into production systems
• Experience in training large-scale models on GPU/TPU clusters