Researcher, World Models

Auxo Talent; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Job summary

Auxo Talent is seeking a Researcher to advance world models for humanoid robotics in the Bay Area. The role involves designing, training and evaluating models that let the robot perceive, predict and act across visual, proprioceptive, and force/torque data, in close collaboration with platform, firmware, and hardware teams.

This early-career position emphasizes genuine ownership beyond narrow scope, with opportunities to publish or open-source work and contribute to sim-to-real transfer and

Qualifications

• A proven modelling track record with trained models and solid evaluations.
• JEPA fluency and ability to reason about joint-embedding predictive approaches.
• Breadth across approaches, including familiarity with vision-language-action models and trade-offs.
• Depth in at least one sensory modality: vision, audio, language, or similar.
• Strong data abilities and ability to work without a dedicated data-engineering team.
• Solid engineering skills to implement and ship research

Responsibilities

• Design, train and evaluate world models for predicting action consequences across modalities.
• Advance self-supervised learning stacks for visual and sensor representations using JEPA family.
• Prototype and benchmark generative and predictive architectures (diffusion, DiT, VAEs) against JEPA objectives.
• Own end-to-end data pipeline: curation, tooling and scaling.
• Integrate research with platform, firmware and software teams to deploy on robots.
• Contribute to sim-to-real transfer, inverse dynamics and multi-modal sensor fusion; publish/open-source work

Job description

Researcher, World Models (Humanoid Robotics)

Location: Bay Area

About the Role

We're building the world models that let a humanoid robot perceive, predict and act in the real world. We're looking for a Researcher to help advance that core capability, working at the intersection of self-supervised representation learning, predictive architectures and embodied control, in close collaboration with our platform, firmware and hardware teams.

This role suits someone early in their research career, roughly a year or so in, who's ready for genuine ownership rather than a narrow, tightly scoped lane.

What You'll Do

• Design, train and rigorously evaluate world models that let the robot predict the consequences of actions across visual, proprioceptive and force/torque modalities
• Advance our self-supervised learning stack for visual and sensor representations, building on and extending the JEPA family (V-JEPA, I-JEPA and related predictive-embedding approaches)
• Prototype and benchmark generative and predictive architectures (diffusion, DiT, flow matching, VAEs) against JEPA-style objectives for embodied prediction and planning
• Own the data pipeline for your experiments end to end, including curation, tooling and scaling, without depending on a separate data-engineering team to move
• Integrate what you build with our platform, firmware and software teams so your research reaches the robot, not just the paper
• Contribute to sim-to-real transfer, inverse dynamics and multi-modal sensor fusion, and publish or open-source work where it strengthens the field and the team

What We're Looking For

• A proven modelling track record: you've trained models and can show solid, honest evaluations, not just training curves
• JEPA fluency: you understand the joint-embedding predictive approach and can reason about where it fits versus alternatives
• Breadth across approaches, including familiarity with VLA (vision-language-action) models and a view on their trade-offs
• Depth in at least one sensory modality: vision, audio, natural language or similar
• Strong data abilities: you get things done without depending on a whole data-engineering team
• Solid engineering: you can implement, integrate and ship what you build alongside platform, firmware and software teams
• A humanoid robotics background, ideally hands-on, and roughly a year into your research career

Nice to Have

• Publications at NeurIPS, ICML, ICLR, CoRL or RSS (or arXiv work with comparable traction)
• A PhD or equivalent research experience in ML, robotics or computer vision; not required with a strong portfolio
• Demonstrated hardware or robotics interest or hands-on experience
• Strong communication: technical blogs, talks or clear written research