Research Scientist, Robotics RL, DeepMind

Job description

Minimum qualifications

PhD degree in a technical field or equivalent practical experience. 2 years of experiencein reinforcement learning or other techniques for mid-/post-training of foundation models and self-improvement in robotics, and other areas such as multimodal generative modeling, training and inference, vision-language/video, or other multimodal models.

Preferred qualifications

Experience working with simulators and real-world robots, esp. dexterous manipulation including multi-fingered hands or whole body control, as well as multimodal sensing (e.g. tactile). Experience deploying the above models and algorithms on real-world systems. Experience implementing large-scale systems and working with large real world data. A passion for bringing research from the lab to real-world robotic systems. About the job As a part of this role, you will lead our efforts in developing novel algorithms and models for embodied AI and Artificial General Intelligence as we are looking for creative thinkers with strong algorithm-design and programming skills that want to build the next-generation of AI agents for and grounded in the physical world.

You will work in collaborative teams to invent algorithms and innovate on large foundation models such as Gemini Robotics. You will design prototype applications and work with real robots inside and outside the lab to address real-world use cases, strong algorithmic background in scalable machine learning (e.g. reinforcement learning/imitation learning; multimodal foundation models) and experience with real robots/robot simulation and large-scale training setups are valued.

In this role, you will have experience with approaches for mid-/post-training for robotics foundation models, including reinforcement learning and related techniques that enable agents to self-improve and achieve mastery, including experience with the use of Reinforcement Learning (RL) to train robot policies in simulation and in the real world. You will have experience with such techniques in the context of general multimodal frontier models (e.g.

LLMs, VLMs, video models) and agentic systems are also highly valued.