Senior AI Researcher

About the role

We are building next-generation world models that enable robots to learn, plan, and act through imagined futures. As a Senior Research Scientist, you will design and scale action-conditioned models that serve as reliable foundations for policy learning, control, and real-world deployment. You will work at the intersection of generative modeling, dynamics learning, and robotics, with the goal of making learned simulators useful—not just predictive—for real-world decision-making.

You will take ownership of significant research problems, make important architectural decisions, anticipate modeling and scaling risks, and collaborate closely with founders, product leaders, and other researchers.

Responsibilities

• Design world models for prediction and control
• Design action-conditioned world models with expressive latent representations, stable rollouts, and control-oriented predictions.
• Improve long-horizon fidelity under autoregressive use, not only one-step prediction accuracy.
• Integrate video priors, physical structure, and object-centric representations into learned control systems.
• Explore latent-action interfaces for cross-embodiment transfer, including human-to-robot transfer.
• Build closed-loop learning systems
• Develop methods for policy learning inside learned simulators, including actor-critic learning over imagined trajectories.
• Build closed-loop training pipelines in which models and policies co-evolve.
• Develop systems that bridge simulation and reality through digital twins, online adaptation, or related approaches.
• Evaluate trade-offs across fidelity, robustness, latency, and inference cost in real robotic settings.

Requirements

• Strong background in machine learning, computer vision, robotics, or a related field.
• Hands-on experience designing and training generative models, including diffusion models, autoregressive video models, or related sequence architectures.
• Experience with at least one of the following: World models or learned dynamics models, Generative video modeling, Model-based reinforcement learning or planning, Robot learning or embodied AI, System identification, physics-informed learning, or simulation.
• Strong understanding of long-horizon prediction, autoregressive rollout, and the failure modes that emerge when models operate on their own outputs.
• Experience working across model architecture, training systems, experimentation, and evaluation.
• Able to take ambiguous research problems from first principles through implementation.
• Strong technical judgement and experience making consequential modeling or architectural decisions.
• Able to communicate research direction clearly and collaborate effectively across research, engineering, product, and leadership.

Nice-to-Haves

• Experience training policies inside learned simulators or imagined trajectories.
• Experience with action-conditioned video prediction or controllable generative models.
• Experience connecting learned models to real robotic systems.
• Familiarity with latent-action models, cross-embodiment learning, or learning from human video.
• Experience with object-centric representations, physical priors, or structured dynamics models.
• Experience with digital twins, sim-to-real transfer, online adaptation, or closed-loop data collection.
• Experience scaling research systems across large datasets or distributed training environments.
• Publications at leading machine-learning, computer-vision, or robotics venues.
• PhD or MS in Computer Science, Machine Learning, Robotics, or a related field.