Sareh Nabi

I am an AI Researcher at Stanford University. My research seeks to deepen our understanding of fundamental principles across the natural sciences and artificial intelligence, using reinforcement learning (RL), multi-agent RL, deep learning, and large language models (LLMs), and to apply these techniques to complex real-world problems. My current projects study emergent behavior in societies of AI agents, neuro-symbolic architectures that integrate LLMs with formal logic, and knowledge-augmented theorem proving in formal mathematics.

I was at Amazon from 2021 to 2026. I completed my postdoctoral research there in September 2023 under Dr. Lihong Li's supervision as part of Amazon's Early Career Scientist Program, focusing on multi-agent RL and LLMs for Amazon's advertising solutions. I then applied LLMs to Amazon's services, including customer service chatbots, with contributions in reasoning, intent understanding, real-time mitigation, and multi-turn automated evaluation. Prior to that, from 2018 to 2021, I was an Applied Scientist at Microsoft, where I built machine learning solutions for enterprise business applications.

I earned my Ph.D. in Operations Research from the University of Washington in 2018, advised by Prof. Hamed Mamani and co-advised by Dr. Houssam Nassif. My doctoral research focused on applications of contextual bandits in dynamic pricing and online advertising. I hold master's degrees in Mathematics and Economics from Simon Fraser University in Canada and an undergraduate degree in Mathematics from Sharif University of Technology in Iran. My master's research focused on modeling biological aggregations using partial differential equations.

My CV with more details here.