Research Engineer, Gemini API and Beyond Serving Compute Efficiency, Deepmind

Google Google · Big Tech · Mountain View, CA +2

Research Engineer role focused on improving compute efficiency for AI serving infrastructure, specifically for platforms like AI Studio and Gemini API. The role involves researching and developing quantitative approaches to optimize fleet-wide compute usage, reporting on costs and utilization, and contributing to analytical infrastructure. Collaboration with product managers, engineers, and researchers is key, with a focus on informing strategy and deploying scalable AI solutions.

What you'd actually do

  1. Lead the reporting of compute costs and utilization, while implementing critical business, product, and engineering metrics for platforms like AI Studio, Gemini API, and Beyond.
  2. Research and develop quantitative approaches to improve fleet-wide compute efficiency, and connect these usage patterns to product and business strategy.
  3. Contribute to innovative reporting and analytical infrastructure, including the use of agentic workflows and advanced dashboarding techniques.
  4. Present and communicate complex research findings in a compelling manner to a varied set of stakeholders.
  5. Collaborate across DeepMind (GDM) organizations to support high-quality, cross-product journeys for developers.

Skills

Required

  • Python
  • C++
  • statistical methods
  • experimental ideas
  • product team experience

Nice to have

  • Master's degree or PhD/DPhil
  • large visual, language, and multimodal research
  • developer experience
  • generative AI
  • front-end design skill
  • compute allocation
  • serving architecture
  • data visualization
  • cross-functional collaboration

What the JD emphasized

  • critical business, product, and engineering metrics
  • improve fleet-wide compute efficiency
  • agentic workflows
  • productionizing state-of-the-art large visual, language, and multimodal research
  • compute allocation and serving architecture

Other signals

  • compute efficiency
  • AI Studio
  • Gemini API
  • DeepMind