Machine Learning Engineer

Machine Learning Engineer — World Models

We are building large-scale neural simulators that model the dynamics of the physical world, enabling robots to understand and anticipate events, so they can plan actions instead of reacting. The focus is on extending video generation systems into

controllable, persistent world models

capable of reasoning over time, causality, and interaction.

What You’ll Work On

• Designing spatiotemporal architectures for long-horizon prediction
• Training multi-billion parameter models on distributed GPU clusters
• Owning the end-to-end training loop (data, optimization, evaluation, iteration)
• Improving model behavior through failure analysis, data curation, and scaling strategies

Requirements

• Strong coding skills (Python, C\+\+, or Rust)
• Hands-on experience with video generation, diffusion, or multimodal models
• Deep understanding of model architecture and scaling laws
• Experience running and debugging large-scale distributed training jobs
• Comfortable working at the intersection of modeling and systems

Nice to Have

• Work on world models, simulation, or model-based RL
• Experience with spatiotemporal transformers or latent dynamics models
• Familiarity with GPU optimization / distributed training frameworks

This is a highly technical, IC-heavy role focused on pushing beyond short-form generation into structured, long-horizon world modeling.