Research Scientist, Robotics RL, DeepMind

Overview

In this role you will design, train, and evaluate scalable algorithms for robotic agents, focusing on post-training self-improvement and mastery. You will contribute to robotics foundation models and physical agents, collaborating with cross-functional teams to build robust research pipelines and fast-turn experiments. You will develop novel algorithms and test hypotheses on real robots and simulators, translating research into real-world robotic systems. You will help shape the next generation of embodied AI, emphasizing scalable learning, multimodal models, and safe deployment in real-world contexts.

Responsibilities

• Design scalable RL and related techniques for post-training self-improvement of robotics agents
• Develop and evaluate algorithms for embodied AI and multimodal foundation models
• Build robust software pipelines to run rapid experiments and test hypotheses
• Collaborate with cross-functional teams to translate research into real-world robotic systems
• Prototype applications and test on real robots and simulators
• Communicate findings clearly to internal and external audiences

Key requirements

• PhD in a technical field or equivalent practical experience
• 2 years of experience in reinforcement learning or post-training methods for robotics and foundation models
• Experience with simulators and real-world robots, dexterous manipulation preferred
• Experience deploying models and algorithms on real-world systems
• Strong algorithmic background in scalable ML (RL/IL; multimodal models) and large-scale training setups

• creativity
• collaboration
• communication
• reinforcement learning
• imitation learning
• multimodal foundation models

…