Mingyang (Michael) Zhang

I lead LLM personalization and search tool use at the Meta Superintelligence Lab. My research work spans from LLM foundation model post-training, LLM context & harness engineering, to LLM agents & applications.

Spearheaded and directed a research team at Google DeepMind focused on improving Large Language Model (LLM) user intent understanding and personalization. My research topics spanned from LLM context retrieval (e.g. RAG, long context LLM), LLM foundation model post-training & alignment, to LLM applications. My research outcomes directly fueled the advancement of the Gemini models, the Gemini App, Google Search AI Overview and AI Mode.

I led a research team at Google Research focused on natural language understanding, information retrieval, and discovery, with a special emphasis on personalized large language models (LLMs). We pushed the boundaries of state-of-the-art technologies and successfully deployed our research in the Gemini models as well as major Google products such as Search, Ads, Maps, and Workspace. I joined Google in 2013 as an L4 Software Engineer and was gradually promoted to Senior Staff Software Engineer and Senior Manager over the years.