Research Scientist – World Models, Robotics & Embodied AI

Meta Meta · Big Tech · London, United Kingdom

Research Scientist role focused on developing AI and robotics technology, specifically Generative World Models for robotic agents to perceive, understand, reason about, and interact with the world. The role involves algorithm development for predictive world modeling, active perception, and robotic interaction, spanning from foundational models to action in the physical world.

What you'd actually do

  1. Drive fundamental and applied research at the intersection of multi-modal generative AI, predictive world modeling, embodied reasoning, and robotic manipulation
  2. Investigate or invent architectures that deliver a spectrum of embodied behaviors from simulated environments to real robots, and from tactile-driven motor control to high-level, long-horizon intelligence
  3. Design research methodologies and lead empirical evaluations, authoring well-tested code for physical hardware and simulators
  4. Build prototype systems that drive multi-step, long-horizon robotic perception, reasoning, and action
  5. Contribute to and lead high-impact publications and open-sourcing efforts

Skills

Required

  • PhD in Computer Vision, Robotics, AI, or related field, or equivalent practical experience
  • Deep learning frameworks (e.g., Pytorch, Tensorflow)
  • Python
  • Experience with real-world robots and simulators
  • Experience with 3D computer vision algorithms
  • Experience training/evaluating 3D generative models, world models, or foundational AI models for embodied tasks
  • Experience with Vision-Language-Action models (VLAs)
  • Reinforcement Learning (RL)
  • Long-horizon planning
  • Kinematics
  • Sim2Real transfer
  • Demonstrated research and software engineering experience

Nice to have

  • multi-modal Generative World Models
  • predictive world modeling
  • embodied reasoning
  • robotic manipulation
  • active perception
  • robotic interaction
  • tactile-driven motor control
  • high-level, long-horizon intelligence
  • physical hardware and simulators
  • multi-step, long-horizon robotic perception, reasoning, and action
  • open-sourcing efforts
  • foundational AI models for embodied tasks

What the JD emphasized

  • PhD in the field of Computer Vision, Robotics, AI, Computer Science, a related field, or equivalent practical experience
  • Experience with real-world robots and simulators
  • Track record of results as demonstrated by grants, fellowships, patents, as well as publications at peer-reviewed workshops, journals, or conferences such as CVPR, CoRL, ICRA, RSS, NeurIPS, ECCV, ICCV, IROS, or similar

Other signals

  • multi-modal Generative World Models
  • robotic manipulation
  • embodied reasoning
  • real-world autonomous applications
  • egocentric devices
  • physical robotic platforms