Fan YANG nameimg


Researcher
Systems and Networking Group
Microsoft Research Asia

Personal Contact: yang DOT fan AT 163 DOT com
Fan

About Me

I am currently a systems researcher in the Systems and Networking Research Group at Microsoft Research Asia (MSR-Asia). I joined MSR-Asia in 2004, after receiving my Ph.D. in the department of computer science at Nanjing University, 2003.

My recent research focuses more on machine learning system, including new challenges in large-scale compute cluster for AI, Deep Learning compiler, and AutoML toolkit. I am an architect of the machine learning systems OpenPAI, NNFusion, and NNI, which have been used by Microsoft products like Bing and Azure. As open-source projects, these systems also have a growing number of external users across the world.

In the past, I worked on graph systems. The graph systems I co-developed like GraM set a new speed record for trillion-scale graph analytics and have been used by Bing to improve Ads coverage and web mining. I also contributed to SCOPE, Microsoft’s big data engine. I helped re-architect and optimize the SCOPE-based analytic pipeline of Bing Ads team, improving its processing capacity by more than 50%, which generated tens of millions of dollar of additional revenue.

Some Tips on Systems Research

During the research and development of complex computer systems, I found that Richard Gabriel’s "worse-is-better philosophy" quite helpful, especially the following: “It is slightly better to be simple than correct.” I would love to hear more real-world examples following this philosophy.

Systems research is built on a series of seminal works, SIGOPS Hall of Fame Award covers most of such works. And if you are looking for advice on general research methodology, Richard Hamming’s talk on “You and Your Research” is a must-read.