Senior Product Manager, Gemini Post-training, Deepmind

Google Google · Big Tech · Mountain View, CA +1

Product Manager for Gemini Post-Training, focusing on integrating RL environments across professional verticals for long-trajectory actions, multi-step tool use on real-world APIs, and automatic context management. The role involves defining user journeys, measuring model generalization, and designing post-training evaluations for Gemini's quality in world knowledge, tool use, and context engineering.

What you'd actually do

  1. Oversee the generation and integration of data across many expert domains targeting production-level quality for hours-long agentic tasks. Define key user journeys and relevant task distributions on the basis of User Experience Research (UXR), user interviews, and engaged industry analyses.
  2. Work closely with modeling teams to carefully measure and improve model generalization to real Application Programming Interface (APIs).
  3. Design new internal post-training evaluations for the most critical areas, demonstrating Gemini quality across world knowledge, tool use, and context engineering.

Skills

Required

  • product management
  • AI/ML
  • data systems
  • agentic frameworks
  • technical field
  • User Experience Research (UXR)
  • user interviews
  • industry analyses

Nice to have

  • managing large-scale data pipelines
  • dataset curation
  • environment generation for machine learning
  • lead cross-functional teams
  • shape research culture
  • manage complex stakeholder relationships
  • highly ambiguous environments

What the JD emphasized

  • long-trajectory actions (hours-long tasks)
  • multi-step tool use on real-world APIs
  • automatic context management
  • post-training evaluations

Other signals

  • orchestrate a massive domain expansion
  • integrating Reinforcement Learning (RL) environments
  • long-trajectory actions (hours-long tasks)
  • multi-step tool use on real-world APIs
  • automatic context management
  • translating these wins to first- and third-party harnesses
  • land these capabilities to millions of users
  • post-training evaluations
  • Gemini quality across world knowledge, tool use, and context engineering