Research Scientist, Strategic Bets, Deepmind

Google Google · Big Tech · London, United Kingdom

Research Scientist at Google DeepMind focusing on building AI systems for probabilistic estimation and forecasting. The role involves designing novel reasoning approaches, multi-agent workflows, and addressing core forecasting challenges. It requires collaboration with researchers, engineers, and product teams for deployment, with an emphasis on publishing research findings. Minimum qualifications include a PhD in AI/ML, experience with LLMs, agentic workflows, and probabilistic modeling. Preferred qualifications include experience with prediction markets and RAG.

What you'd actually do

  1. Engage with existing research efforts to build AI systems capable of superhuman probabilistic estimation, focusing on domains with no clean time series or stable reference classes.
  2. Collaborate with researchers and engineers to design novel structured reasoning approaches, iterative self-improvement loops, and multi-agent workflows that improve estimation quality and calibration.
  3. Research solutions to core forecasting issues, including temporal reasoning, information retrieval and filtering, coherent world-modeling, and novel reward structures for multi-step reasoning.
  4. Engage with product teams and enterprise partners to drive the deployment of our research into high-stakes decision-making environments.
  5. Report and present research findings clearly and efficiently both internally and externally, and suggest team collaborations to meet research goals for the wider Science program.

Skills

Required

  • PhD degree in Machine learning, AI, or a related computational field, or equivalent practical experience.
  • Experience in delivering research impact through publications, open-source contributions, or deployed systems.
  • Experience with large language models, agentic workflows (e.g., tool use, decomposition, multi-agent coordination), or inference-time reasoning architectures.
  • Programming experience across common scripting languages and ML pipelining tools.
  • Experience in probabilistic modeling, uncertainty quantification, calibration, or reinforcement learning.

Nice to have

  • Experience with prediction markets, quantitative research, temporal data modeling, or real-world forecasting benchmarks.
  • Familiarity with process-based reward modeling, automated design of agentic systems, or search/retrieval-augmented generation.

What the JD emphasized

  • Experience in delivering research impact through publications, open-source contributions, or deployed systems.
  • Experience with large language models, agentic workflows (e.g., tool use, decomposition, multi-agent coordination), or inference-time reasoning architectures.
  • Experience in probabilistic modeling, uncertainty quantification, calibration, or reinforcement learning.

Other signals

  • AI systems capable of superhuman probabilistic estimation
  • novel structured reasoning approaches
  • multi-agent workflows
  • core forecasting issues, including temporal reasoning, information retrieval and filtering, coherent world-modeling, and novel reward structures for multi-step reasoning