Research Scientist, Robotics

<h2>Snapshot</h2> <p><span style="font-weight: 400;">Artificial Intelligence could be one of humanity’s most useful inventions. At Google DeepMind, we’re a team of scientists, engineers, machine learning experts and more, working together to advance the state of the art in artificial intelligence. We use our technologies for widespread public benefit and scientific discovery, and collaborate with others on critical challenges, ensuring safety and ethics are the highest priority.</span></p> <h2>About Us</h2> <p>At Google DeepMind Robotics we are pioneering bringing AI to the physical world. We are powering an era of physical agents, enabling robots to perceive, plan, think, use-tools, and act to better solve complex tasks.&nbsp;</p> <div>What we do:</div> <div> <ul> <li><strong>Building foundation models:</strong>&nbsp;We develop advanced Vision Language Action (VLA) models that combine Gemini's world understanding with physical actions to directly control robots. These models include Gemini Robotics (our most advanced Gemini model for the physical world), and Gemini Robotics On-Device (our fastest Gemini model that runs without a data network). We also develop advanced reasoning and agentic systems for the physical world, including: Gemini Robotics-ER, a Gemini agent with advanced spatial understanding and agentic reasoning. These models enable robots to perform a broad range of tasks, respond interactively to their environment, achieve high dexterity, and reason over long multi-step tasks.</li> <li><span class="im"><strong>Advance the state of the art:</strong>&nbsp;We are always focused on advancing the state-of-the-art in all areas of general purpose robotics, including agentic reasoning, real world understanding, action generalization, human robot interaction, dexterity, whole-body-control, continual learning, and more.</span></li> <li><span class="im"><strong>Deploy at scale:</strong> We partner with key robotics companies to bring this intelligence to the physical world across a broad range of applications at scale.</span></li> </ul> </div> <h2>The Role</h2> <p>Research Scientists lead our efforts in developing novel algorithms and models for embodied AI and Artificial General Intelligence. 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. Research Scientists work in collaborative teams to invent algorithms and innovate on large foundation models such as Gemini Robotics. They design prototype applications and work with real robots inside and outside the lab to tackle real-world use cases. A 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.</p> <h3>Key responsibilities:</h3> <ul> <li>Design, implement, train and evaluate large models and algorithms for robotic agents. Make breakthroughs and unlock new robot capabilities.</li> <li>Write software to implement research ideas and iterate quickly.</li> <li>Leverage your expertise to participate in a wide variety of research: learning from simulation, reinforcement learning, learning from demonstrations, vision-language-action models, transformers, video generation, robot control, humanoid robots and more.</li> <li>Work effectively with a large collaborative team with fast-paced agendas to meet ambitious research goals.</li> <li>Generate creative ideas, set up experiments and test hypotheses. Report and present research findings clearly and efficiently both internally and externally.</li> </ul> <h2>About You</h2> <p><strong>In order to set you up for success as a Research Scientist at Google DeepMind, &nbsp;we look for the following skills and experience:</strong></p> <ul> <li>PhD in a technical field or equivalent practical experience.</li> <li>A strong algorithmic background with deep knowledge of areas such as&nbsp;</li> <ul> <li>reinforcement learning and imitation learning</li> <li>multimodal generative modeling, training and inference.</li> <li>large vision / vision-language / video and other multimodal models</li> </ul> </ul> <p><strong>In addition, the following would be an advantage:&nbsp;</strong></p> <ul> <li>Experience working with simulators and real-world robots, esp. dexterous manipulation including multi-fingered hands and / or