Research Scientist - Embodied World Models
Core
Design, train, and evaluate world models that predict action outcomes for humanoid robots in simulation and real-world scenarios.
Role type
Senior IC research scientist (embodied AI/world models)
Builds
Self-supervised world models for robot manipulation and locomotion
Domain
Robotics, physical AI, self-supervised learning
Deliverable
production ML models
Required skills
robotics, representation learning, world models, self-supervised learning, RL/planning, video modeling, large-scale training, efficient inference, fast experimentation, end-to-end ownership
Preferred skills
latent world models (JEPA), early-stage startup experience, core team experience at big tech building humanoids
Technologies
simulation environments, real-world robot hardware, control stacks
Responsibilities
Design and analyze original experiments on world models and planning; own the stack from data processing to real-robot evaluation; push self-supervised pretraining across video and action data; integrate research with hardware and control stacks; mentor engineers
Seniority
Senior, hands-on IC with research originality and engineering pragmatism
