James Cheng

James Cheng is a computer scientist at the Chinese University of Hong Kong whose research grew from graph data management and distributed data systems into machine learning systems and LLM agents. He was principal investigator of an NTU start-up grant on large-scale graph data from 2009 to 2012 and received the Hong Kong Young Scientist Award in 2010. A 2014 paper with his student Yanyan Xu won the Best Student Paper Award at IEEE BigData, and the ParSync cluster scheduler, written with Alibaba Group, won Best Paper at USENIX ATC in 2021 and runs in Alibaba's Fuxi. In 2024 he coauthored COAT, a NeurIPS paper that uses large language models to find hidden causal factors. He is a professor of computer science and engineering at CUHK, where his Husky Data Lab works on LLM agents, reinforcement learning and multimodal foundation models.

Now at: Chinese University of Hong Kong

Timeline

June 2009–May 2012

NTU start-up grant on graphs

He serves as principal investigator on an NTU Start-Up Grant of SGD 100,000 from June 2009 to May 2012. The project, Large-Scale Graph Data Management, covers graph data mining, graph pattern matching and retrieval, graph indexing and graph data storage.
2010

Hong Kong Young Scientist Award

He receives the Hong Kong Young Scientist Award in Physical/Mathematical Science in 2010.
2013–2020

Teaching database systems at CUHK

From 2013 he teaches Advanced Topics in Database Systems in the CUHK computer science and engineering department, and the course returns in 2015, 2018 and 2020. In 2016, 2017 and 2018 he also teaches Large Scale Data Processing Systems, covering distributed data analytics, graph processing and stream processing systems.
2014

Best student paper at BigData

In 2014, his paper with student Yanyan Xu, Distributed Maximal Clique Computation, wins the Best Student Paper Award of the IEEE International Congress on Big Data.
2021

ParSync wins Best Paper

In 2021, he coauthors Scaling Large Production Clusters with Partitioned Synchronization with Alibaba Group colleagues, which wins Best Paper at USENIX ATC. The ParSync scheduler it describes is deployed in Fuxi, the distributed cluster scheduler at Alibaba.
2024

COAT at NeurIPS

In 2024, he coauthors COAT, a NeurIPS 2024 paper with Chenxi Liu, Yongqiang Chen and others. COAT uses large language models to propose causal factors from unstructured data, which a causal learning module then tests.
As of October 2026

Professor at CUHK

He is a professor in the Department of Computer Science and Engineering at The Chinese University of Hong Kong. His research covers evaluation and rubric-based reward design for LLM agents, reinforcement learning for long-horizon reasoning, and multimodal and omni foundation models. His Husky Data Lab has built the Husky distributed computing system, among other large-scale systems.