In 1986, Kai Li published his PhD dissertation entitled "Shared Virtual Memory on Loosely Coupled Microprocessors", thus opening up the field of research that is now known as distributed shared memory which allows users to program using a shared-memory programming model on clusters. Since this work, there has been a huge amount of work done to extend the idea to other areas and to improve DSM's performance. After joining Princeton, Li himself also led the Scalable High-performance Really Inexpensive MultiProcessor project, which investigated how to build high-performance servers on a cluster. During his Princeton career, Li co-led the Scalable I/O project which attacked I/O bottleneck problems for supercomputers. His work with protected user-level communication has contributed significantly to the Remote Direct Memory Access mechanism, the Virtual Interface Architecture standard, and the Infiniband standard, which are the communication mechanisms for the Direct Access File System. Li also led the Scalable Display Wall project, which explores how to build and use a high-resolution, wall-size display system to visualize massive datasets. Recently, he has been working with colleagues at Stanford on a large well-labelled image dataset called ImageNet to help computer vision community develop object recognition and classification methods for large image data. More recently, he has been working with colleagues at Princeton on developing methods to efficiently manage and analyze the vast amount of data generated by increasingly sophisticated research in fields ranging from genomics to neuroscience. His google scholar statistic shows his prime contributions in virtual memory and data storage.
Entrepreneur experience
During his sabbatical from Princeton in 2001, with Ben Zhu and Brian Biles, Li co-founded Data Domain Corporation which built the first commercial deduplication storage system, opening up a new billion-dollar market. In June 2009, EMC Corporation acquired Data Domain for $2.4 billion, outbidding NetApp's previous offer. In 2010, Data Domain product line captured 64.2% share of the market for purpose-built backup devices worldwide, including mainframes. In Data Domain, he served as initial Chief Executive Officer and then Chief Technology Officer from October 2001 to August 2002. Since September 2002, he has served as its Chief Scientist, being responsible for leading the strategic direction for Data Domain's product vision. Prior to joining Data Domain, he served as an industry consultant of AT&T, Bell Labs, Digital Equipment Corporation, Intel, and NEC. He served as a Director of Pattern Insight Inc. and served as advisory board members of EMC Corporation, Inphi Corporation, Intel research labs, and Open Innovation Center of Samsung Corporation. Since 2016, he served on the Board of Directors for Moqi, a venture-backed company which produces a fingerprint matching system employing cutting-edge large-scale precise image search algorithms, with broad applications in law enforcement, government, healthcare and financial industries. Moqi, whose main client is the Ministry of Public Security, is based in Beijing and was founded by Cheng Tai and Linpeng Tang, both PhDs from Princeton University. In September 2019, Kai Li visited the company, giving a talk about his entrepreneurship experience.
