Research Engineer - Embodied World Models
Core
Building world models that predict action outcomes to enable robots to understand, plan, and reason in the physical world.
Role type
Research Engineer (Embodied AI / World Models)
Builds
Training, data, and evaluation pipelines for self-supervised learning on video and action data; experiments in latent-space planning and model-predictive control.
Domain
Robotics, Embodied AI, Self-supervised Learning
Deliverable
production ML models
Required skills
Python, PyTorch, distributed/GPU training, self-supervised representation learning, software engineering
Preferred skills
video models, reinforcement learning, planning, robotics, world models (e.g. JEPA), early-stage startup experience, humanoid robot development
Technologies
Python, PyTorch
Responsibilities
Implement and reproduce state-of-the-art world-model baselines; build and own training/data/evaluation pipelines; execute latent-space planning and model-predictive control experiments on real hardware and simulation; profile, debug, and scale distributed training runs.
Seniority
Mid-to-Senior, hands-on IC
