Embodied AI Robotics Scientist — RL & Multimodal Models

DeepMind Technologies Limited; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

DeepMind Technologies Limited seeks researchers to advance embodied AI and robotics foundation models. You will design and prototype algorithms, test them on real robots and in simulation, and collaborate across teams to push the boundaries of AI grounded in the physical world.

You will work on large foundation models and mid-/post-training techniques to enable robot mastery, with emphasis on scalable ML, reinforcement learning, and multimodal systems.

Qualifications

• PhD in a technical field or equivalent practical experience.
• 2 years of RL or related mid-/post-training techniques for foundation models in robotics and multimodal domains.

Responsibilities

• Design, train, and evaluate scalable robotic algorithms focusing on post-training, self-improvement and mastery.
• Contribute to robotics foundation models and physical agents research themes, including VLAs, WAMs and imitation learning.
• Develop scalable research pipelines and robust software to test hypotheses quickly.
• Collaborate in a fast-paced team to pursue ambitious research goals.
• Generate ideas, set up experiments, and clearly report findings internally and externally.

Skills

Reinforcement learning

Robotics

Algorithm design

Programming

Multimodal models

Education

PhD or equivalent

Tools

Simulation frameworks

Job description

DeepMind Technologies Limited seeks researchers to advance embodied AI and robotics foundation models. You will design and prototype algorithms, test them on real robots and in simulation, and collaborate across teams to push the boundaries of AI grounded in the physical world.

You will work on large foundation models and mid-/post-training techniques to enable robot mastery, with emphasis on scalable ML, reinforcement learning, and multimodal systems.