Rujun Han
Senior Research Scientist at Google
I work on LLM Agents and post-training at Google. My recent work on Deep Research is now available on Gemini Enterprise and Gemini DR Max with millions of users. Previously, I spent two years at AWS AI Labs leading research in RAG, LLM evaluations, and alignment while building enterprise AI assistant services Kendra and Amazon Q.
My current research spans two complementary directions:
- Vertical: I build agents that solve challenging enterprise use cases in finance, engineering design, supply chain, and more.
- Horizontal: I invent general agent capabilities that enable agents to curate skills and learn from their own experience via recursive self-improvement.
Timeline
2024 – present
Senior Research Scientist
Google — Cloud AI Research
Lead agentic research and products.
2022 – 2024
Applied Scientist
Amazon Web Services — AI Labs
Led research and products in agentic RAG applications.
2017 – 2018
Senior Associate, Data Scientist
Commonwealth Bank
Built the bank's first ML model for lending risk.
2015 – 2017
Senior Research Analyst
Federal Reserve Bank of New York
Built economic forecasting models, including Nowcast.
2018 – 2022
Ph.D. in Computer Science
University of Southern California
2015 – 2018
M.Sc. in Data Science
New York University
2010 – 2014
B.A. in Economics and Mathematics
Hamilton College
Selected Publications
See all publications on Google Scholar.
AI for Physical Engineering
Agent Harness: Memory, Skills, Recursive Self-Improvement (RSI)
Deep Research and Information Retrieval (RAG) Agents
Powers Gemini Enterprise Deep Research
EACL'26 Findings
Post-training: RL, Distillation
ICLR'25