ORC IAP Seminar 2026 Talk 4 Sirui Li
Sirui Li
Senior Research Engineer
Microsoft Research
Title
OptiGuide: GenAI for Optimization and Decision Intelligence
Abstract
In this talk, we describe how large language models (LLMs) can be leveraged to improve optimization. We present applications to Microsoft’s cloud supply chain, where LLMs provide insights to planners and support what-if analysis. We then discuss recent research on improving the fine-tuning performance of LLMs for optimization, including the use of LLMs for optimization data synthesis and a domain-specific, semi-automated pipeline for improving synthetic optimization data quality. We show that post-training competitive open-source models such as gpt-oss-20b on this data leads to substantial gains on optimization formulation tasks, and we conclude by outlining future research directions at the intersection of operations research and AI.
Bio
Sirui Li is a Senior Research Engineer at Microsoft Research in the Machine Learning and Optimization (MLO) group. Her current research focuses on the intersection of large language models and optimization, aiming to improve optimization to be more accessible, interpretable, and reliable. She finished her PhD in Social and Engineering Systems (SES) and Statistics (IDPS) at MIT, where she was advised by Prof. Cathy Wu on machine learning for combinatorial optimization. Her doctoral research developed learning-guided algorithms to accelerate large-scale combinatorial optimization solvers, as well as control-theoretic and reinforcement learning methods for mixed-autonomy transportation systems.