World Models Researcher — Humanoid Robotics & SSL

Menlo; United States; Remote; USD 120,000 - 180,000; Full time

Job summary

the company Research is seeking a Researcher to advance world models central to Asimov’s perceptual, predictive, and active capabilities. You will work at the intersection of self-supervised representation learning, predictive architectures, and embodied control, collaborating with platform, firmware, and hardware teams.

You will design and evaluate world models, advance self-supervised stacks, prototype revolutionary architectures, and own end-to-end data workflows, ensuring the research

Qualifications

• Proven modeling track record.
• JEPA fluency and ability to reason about joint-embedding predictive approaches.
• Breadth across vision-language-action models and trade-offs.
• Depth in at least one modality (vision, audio, language, etc.).
• Strong data capabilities and ability to operate without a full data-engineering team.
• Solid engineering; able to implement, integrate, and ship research with platform teams.

Responsibilities

• Design, train, and evaluate world models for predicting action consequences across visual, proprioceptive, and force/torque modalities.
• Advance self-supervised learning for visual and sensor representations.
• Prototype and benchmark generative/predictive architectures (diffusion, DiT, flow matching, VAEs) against JEPA objectives.
• Own the data pipeline end-to-end: curation, tooling, scaling.
• Integrate research outputs with platform, firmware, and software teams to deploy to the robot.
• Contribute to sim-to-real transfer, inverse dynamics, and publish/open-source work as appropriate.

Skills

World models

Self-supervised

JEPA

Vision fusion

PyTorch

JAX

Tools

Diffusion

DiT

Flow matching

VAEs

Job description

the company Research is seeking a Researcher to advance world models central to Asimov’s perceptual, predictive, and active capabilities. You will work at the intersection of self-supervised representation learning, predictive architectures, and embodied control, collaborating with platform, firmware, and hardware teams.

You will design and evaluate world models, advance self-supervised stacks, prototype revolutionary architectures, and own end-to-end data workflows, ensuring the research