Research Scientist, Robotics Data, Pretraining, DeepMind

Google•Mountain View, CA

3d•$147,000 - $210,000

About The Position

The Research Scientist will help combine Gemini foundation models with physical intelligence so general-purpose robots can understand and act reliably in the real world. The role focuses on obtaining high-quality training data and integrating it into Gemini pre-training. The team develops insights into extracting physical and action knowledge from robot-related datasets and applies them through model training and upstreaming. By pairing data ablations with fast closed-loop robot evaluations, the scientist will help shape physical reasoning and action capabilities in next-generation Gemini models and deploy them onto real-world robots. More broadly, Google Research Scientists set up large-scale tests, rapidly deploy promising ideas, manage deadlines and deliverables, and apply current theories to develop products, processes, or technologies. They create experiments, prototype implementations, and design architectures to address real-world computer science problems. They also contribute to the wider research community by sharing and publishing findings, including work inspired by internal projects and collaborations with universities and technical institutes.

Requirements

PhD in Computer Science, a related technical field, or equivalent practical experience.
Experience working with robot policy learning, such as imitation learning or reinforcement learning, across simulators and real-world robotic platforms.
Experience working with simulators and robots.
Experience contributing to research communities or efforts, including publishing papers at conferences such as NeurIPS, CoRL, ICML, or ICLR.

Nice To Haves

Experience with systematic benchmarking, multimodal data sources, data labeling strategies, and data mixture optimization for frontier or robotics foundation models.
Practical experience iteratively improving model performance through data curation, training recipes, and model benchmarks.
Proven ability to build and maintain clean research software and tools for rapid iteration on research ideas.
Passion for bringing foundation model research from the lab to robust, real-world robotic systems.

Responsibilities

Develop data quality, filtering, and labeling methodologies, and optimize multimodal data representations and mixtures to upstream physical and action knowledge into Gemini pre-training.
Design and run pre-training and policy-training ablations, improve training recipes, and systematically improve physical reasoning and action performance against benchmarks.
Build and maintain research software, evaluation tools, and lightweight post-training workflows to iterate quickly on research ideas and verify upstream data quality through offline and closed-loop robot policy evaluations.
Contribute to the wider research agenda, including vision-language-action modeling, world-action models, reinforcement learning, and simulation.
Communicate findings clearly in a fast-paced, collaborative team environment.
Set up large-scale tests, deploy promising ideas, and manage deadlines and deliverables.
Share and publish research findings and contribute to the wider research community.