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 hill-climb physical reasoning and action performance against benchmarks. Build and maintain the research software, evaluation tools, and lightweight post-training workflows required to iterate on research ideas quickly and verify upstream data quality through offline and closed-loop robot policy evaluations.
Contribute to our wider research agenda—including vision-language-action (VLA) modeling, world-action models, reinforcement learning, and simulation—and communicate findings clearly in a fast-paced, collaborative team environment.
Job description provided by Google.
View the original posting
