Machine Learning Engineer - Realtime Games, Digital Optimus

Job Description

Job Responsibilities

• Design and own experiments to improve long-horizon agentic gameplay, with a focus on memory and real-time control
• Lead or make core contributions to large-scale training runs, including data mixture, scaling, infrastructure, and algorithms
• Improve the post-training stack, including scalable environments, rewards, and behavior evaluations
• Design architectures with efficient context management, real-time reactivity, and intelligent planning, capable of independently completing tasks over many days. 
• Develop rigorous benchmarks and techniques to evaluate agent performance in games and measure transfer across different games, embodiments, and general computer use
• Collaborate across research, infrastructure, and production teams to integrate agent systems into Full Self-Driving (FSD) and physical Optimus 
• Iterate rapidly on prototypes, debug complex multimodal ML stacks, and contribute to new research directions

Job Requirements

• Strong technical understanding of all stages of model development, from pre-training to inference
• Demonstrated contributions to at least one large multimodal training run
• Experience with planning and/or memory architectures for long-horizon or agentic systems
• Track record of delivering research from ideation to publication/production
• Hands-on experience with VLMs, VLAs, RL, and large-scale transformers
• A desire to ship models that solve the impossible