Xiao Ma

Xiao Ma is a software engineer at Google DeepMind, focusing on instruction fine-tuning for multimodal models. She co-led the horizontal scaling of reinforcement learning from human and feedback reviews for Gemini 2.5, a method widely adopted by over 400 engineers to launch more than 50 critical features. Her research primarily explores the interface between artificial intelligence and social interaction, with a particular focus on the social implications of trust. Xiao holds a PhD in Information Science from Cornell University, where her research investigated computational understandings of networked trust. She also earned a bachelor's degree in Electronic Engineering from Peking University. In addition to her technical expertise, Xiao is passionate about the intersection of arts and technology, currently serving as a board member for the performing arts venue National Sawdust.

• 2024–Present·Google DeepMind·Software Engineer
  • Responsible for instruction fine-tuning of multimodal models (Gemini), contributing to the development of advanced reasoning and next-generation agent capabilities.
  • Co-led the horizontal scaling of reinforcement learning from human and feedback reviews, enabling over 400 engineers to apply this method across 50+ projects.
• 2018·Facebook Core Data Science·Researcher
  • Investigated issues of trust in social groups.
• 2016·Airbnb·Researcher
  • Researched the relationship between trust and user profiles, enhancing trust on the platform.

• 2014–2019·Cornell University·PhD in Information Science
  • Focus on computational understandings of networked trust, advised by Mor Naaman, Karen Levy, and others.
• 2010–2014·Peking University·Bachelor’s in Electronic Engineering
  • Majored in Electronic Engineering.