Researcher, World Models

Highlights

• Advance world models letting Asimov humanoid predict consequences of actions.
• Your research runs on real hardware with idea-to-walking-robot in weeks.
• Build open-source full-stack humanoid platform turning software into physical labor.

Ideal candidate

A hands-on researcher who trains and honestly evaluates models end-to-end, owns data and integration, and is driven to see predictive learning generalize on real open-source humanoids.

Requirements

• Proven track record training models with solid, honest evaluations
• Fluency in JEPA joint-embedding predictive approach and trade-offs vs alternatives
• Familiarity with VLA models and adjacent robotics ML approaches
• Strong depth in at least one sensory domain such as vision, audio, or language
• Ability to own data curation, tooling, and scaling without data-engineering support
• Solid engineering to implement, integrate, and ship research to real robots
• Conversant in SSL, world models, generative architectures, sensor fusion, PyTorch/JAX and distributed training