Technical Program Manager for Science & Reasoning at DeepMind leads cross-disciplinary engineering programs at the intersection of frontier AI model development and rigorous scientific domains across the life and physical sciences. You will partner with research leads to translate hypotheses into structured roadmaps, data mixtures, and Gemini release milestones, and you will drive end-to-end execution for human data collection, synthetic data generation, SFT, and verifiable RL training environments. You will curate and audit the evaluation suite for scientific reasoning and agentic workflows to ensure benchmarks reflect real-world scientific validity. You will collaborate with post-training and RL leads to analyze checkpoints, inspect reasoning traces, identify capability gaps, and translate findings into training interventions. You will manage cross-functional dependencies across data infrastructure, evaluation platforms, compute allocations, and external domain-expert vendors, synthesizing training dynamics into launch criteria and actionable updates for tech leads. This is an on-site role in Mountain View, California, United States. Compensation for the US region is listed as US: $217000 - $236000 (USD) + 15% bonus target + equity + benefits.

Technical Program Manager for Science & Reasoning at DeepMind leads cross-disciplinary engineering programs at the intersection of frontier AI model development and rigorous scientific domains across the life and physical sciences. You will partner with research leads to translate hypotheses into structured roadmaps, data mixtures, and Gemini release milestones, and you will drive end-to-end execution for human data collection, synthetic data generation, SFT, and verifiable RL training environments. You will curate and audit the evaluation suite for scientific reasoning and agentic workflows to ensure benchmarks reflect real-world scientific validity. You will collaborate with post-training and RL leads to analyze checkpoints, inspect reasoning traces, identify capability gaps, and translate findings into training interventions. You will manage cross-functional dependencies across data infrastructure, evaluation platforms, compute allocations, and external domain-expert vendors, synthesizing training dynamics into launch criteria and actionable updates for tech leads. This is an on-site role in Mountain View, California, United States. Compensation for the US region is listed as US: $217000 - $236000 (USD) + 15% bonus target + equity + benefits.