Research Engineer, Response Quality​/Hermes

Research Engineer, Response Quality/Hermes

Mountain View, California, US

At Google Deep Mind, we value diversity of experience, knowledge, backgrounds and perspectives and harness these qualities to create extraordinary impact. We are committed to equal employment opportunities regardless of sex, race, religion or belief, ethnic or national origin, disability, age, citizenship, marital, domestic or civil partnership status, sexual orientation, gender identity, pregnancy, or related condition (including breastfeeding) or any other basis as protected by applicable law.

If you have a disability or additional need that requires accommodation, please do not hesitate to let us know.

This will be an overview of your team.

At Google Deep Mind, we are a multidisciplinary team of engineers and scientists driving core contributions to Gemini’s post-training evolution. Our mission is to transform Gemini from a "one-shot" answer engine into an adaptive, collaborative partner with rich response formats.

Our work spans the full stack of model alignment and interaction design, including:

• Advanced Modeling Interventions: Developing and scaling SFT, RL, and Reward Modeling (RM) techniques specifically designed to enhance multi-turn reasoning and collaborative behavior.
• Next-Gen Evaluation: Building autonomous user-agent evaluation techniques to simulate complex, multi-turn interactions and quantify the quality of human-AI exchange.
• Rich Multimodal Synthesis: Moving beyond "wall-of-text" responses by integrating images, charts, and interactive widgets to create a more expressive and functional user experience.

Artificial Intelligence could be one of humanity’s most useful inventions. At Google Deep Mind, we’re a team of scientists, engineers, machine learning experts and more, working together to advance the state of the art in artificial intelligence. We use our technologies for widespread public benefit and scientific discovery, and collaborate with others on critical challenges, ensuring safety and ethics are the highest priority.

Overview of the role

As a Research Engineer/Scientist on the Gemini Post-Training team, you will bridge the gap between core model alignment and the next generation of interactive AI. You will be re-architecting how Gemini interacts, reasons, and collaborates with users over complex, multi-turn exchanges.

Your work will involve building high-scale distillation pipelines, developing metrics and evaluation techniques, taking advantage of latest development on agentic framework and autonomous user agents, and designing custom system instructions (SIs) that redefine model behavior, and finding the right training recipes to impact Gemini. This is a high-impact role where your contributions to SFT and Reinforcement Learning will directly shape the multimodal, widget-rich future of the world’s most capable AI.

Add up to 5 responsibilities

• Models:
  Improve model quality for understanding and generation through applying and developing cutting edge modeling interventions, multi-turn reasoning, and rich-format multimodal synthesis.
• Data:
  Unlock new information-seeking and collaborative capabilities through large-scale data distillation and processing, bridging pre-training and post-training.
• Evals:
  Build on top of agentic frameworks to develop better evaluation methods (human, auto-raters, and automated metrics) that measure the quality of open-ended tasks with rich formats.
• Research Execution:
  Drive projects by defining key research questions and designing experiments that provide clear, actionable answers.
• Product Impact:
  Collaborate with cross-functional teams to land research breakthroughs directly into Google products and services.

In order to set you up for success as a Research Scientist at Google Deep Mind, we look for the following skills and experience:

• PhD or MSc in Computer Science, Machine Learning, or a related technical field.
• RS:
  PhD
• Experience in model alignment or post-training techniques, such as SFT, Reinforcement Learning (RL), or Reward Modeling.
• Experience building and maintaining large-scale data processing or distillation pipelines…