Shuaiwen Leon Song

About

Shuaiwen Leon Song is part of Microsoft Deepspeed research group and manages the new Heterogenous Computing Architecture (HCA) group (e.g., Brainwave project). He leads Deepspeed4Science initiative (deepspeed4science.ai) and several other pathfinding projects related to LLMs. He is constantly exploring and exploiting opportunities to make Deepspeed the most efficient large-scale ML system acceleration framework in the world, focusing on producing cutting-edge system R&D that can benefits Azure and Bing users as well as making positive social impact by enabling sophisticated ML system designs for MSR AI4Science and WebXT/Bing 1P customers.

Prior to Microsoft, Shuaiwen was the SOAR associate professor (tenured) at University of Sydney and an adjunct professor at University of Washington. He also held tenure position in U.S. DoE labs. Shuaiwen’s expertise includes design and optimizations for large-scale complex high-performance systems, e.g., accelerator-driven ML systems and compiler. His signature research works include planet-scale virtual reality system design and cross-stack design for efficient and scalable ML systems. His past works in HPC have received several best paper nominations and were featured in U.S. DoE research highlights and other media outlets. He has received IEEE early-career award in high performance computing, IEEE mid-career award in scalable computing, Google Brain faculty award, Meta faculty award, AIR global faculty award and SOAR fellowship. He also won the Australian most innovative engineer award in 2021.

Shuaiwen has published over 85 refereed conference and journal articles. He also served as program track chair, organizing committee or program committee for leading HPC (e.g., Supercomputing) and computer architecture conferences (e.g., ASPLOS, ISCA, HPCA and MICRO). He is an ACM distinguished speaker.