Zhengyan Shi

关于

Zhengyan is a Senior Researcher at Microsoft Research, working on teaching language models (LMs) to code. His research focuses on building learning loops where LMs not only act but also reason inside scalable and self-evolving environments. By letting models plan, converse, and iterate within these realistic sandboxes, he explores how LMs can continually refine themselves through interaction.

He obtained his PhD from University College London. Prior to joining Microsoft, he gained industry experience at Cohere and Amazon. His current research interests include language model post-training and interactive systems.