Zhi Yang is currently an associate researcher in the School of Computer Science. He obtained his Ph.D. from Peking University in 2010. His major research interests include AI computing systems and distributed systems.
Zhi Yang has published more than 60 research papers, and many of them are published in top-tier conferences, such as OSDI, ATC, SIGMOD, VLDB, KDD, NIPS, WWW, ICLR, ICML, ICDCS, SoCC. He has served in the Program Committee of various international conferences including ICDCS, WWW, KDD, DASFAA, ASONAM, VLDB, He won the best student paper award in WWW 2022
His research achievements are summarized as follows:
1)AI computing framework. By re-abstracting the expression layer and hardware layer of AI computing, he proposes a full-stack reconstruction of the underlying system architecture, such as AI compiler RAMMER for neural networks, and the NeuGraph, the first system capable of supporting large-scale graph neural network training.
2)Distributed scheduling and synchronization. He proposed the sharing safety concept in the multi-tenant cluster scheduling, and designed scheduling framework HiveD to realize the theoretical guarantee of safe resource sharing. He also proposed a high-performance heterogeneous distributed training scheme Partial All-Reduce, and a real-time efficient Semi-distributed data synchronization protocol SDPaxos.
3)Sybil detection in social networks. His publication of “Uncovering social network sybils in the wild”shows the assumption of existing graph-based Sybil detection methods can be invalid, which has a significant impact on the social network Sybil detection.
