Hi, my name is Zizhao Mo (莫梓钊). I received my Ph.D. degree in Computer Science from the University of Macau, where I was fortunate to be advised by Prof. Huanle Xu at Cloud and Distributed Systems Lab.
My research interests broadly revolve around optimizing deep learning workloads in real-world systems, such as LLM training/inference and resource scheduling. The optimization goals of my research involve latency, cost, throughput, and energy, etc. I also have a specific interest in designing efficient systems over heterogeneous resources, including CPU-GPU and heterogeneous GPU platforms.
My current research projects:
- LLM workload optimization. Proposing optimization techniques for the LLM inference and training. I am interested in research problems such as latency optimization, throughput improvement, fault tolerance, and energy saving, etc.
- Resource allocation in GPU clusters. Designing fine-grained scheduling policies in the (heterogeneous) GPU cluster to optimize the performance and resource efficiency for deep learning training jobs.
🔥 News
- 2026.08: I will join Great Bay University as an Assistant Professor in September!
- 2026.07: Served as a TPC member of IEEE HPCA’27.
- 2026.05: Served as a TPC member of IEEE/ACM Micro’26.
- 2026.03: I give a talk at Cloudflare in March 3rd.
- 2025.06: 🎉🎉 I defense my Ph.D. thesis!
📝 Publications
ETH: Data-Balanced Heterogeneous Pipeline Parallelism for Training Large Models
- Accepted by NSDI’ 27 (CCF-A, CSRanking) (to appear).
- Authors: Mianjie Yu, Zhongmin Zhao, Zizhao Mo, Jianxiong Liao, Wenxuan Li, Yulin Qiao, Huanle Xu, Wenquan Yang, Yongqiang Yang, Zeren Li, Ruifeng Tang, Chengzhong Xu
Serving Hybrid LLM Loads with SLO Guarantees Using CPU-GPU Attention Piggybacking
- Accepted by Sigmod’ 26 (CCF-A, CSRanking).
- Authors: Zizhao Mo, Junlin Chen, Huanle Xu, Cheng-Zhong Xu
Hetis: Serving LLMs in Heterogeneous GPU Clusters with Fine-grained and Dynamic Parallelism
- Accepted by SC’ 25 (CCF-A, CSRanking).
- Authors: Zizhao Mo, Jianxiong Liao, Huanle Xu, Zhi Zhou, Cheng-Zhong Xu
Fast and Fair Training for Deep Learning in Heterogeneous GPU Clusters
- Accepted by ICS’ 25 (CCF-B, CSRanking).
- Authors: Zizhao Mo, Huanle Xu, Wing Cheong Lau
Heet: Accelerating Elastic Training in Heterogeneous Deep Learning Clusters
- Accepted by ASPLOS’ 24 (CCF-A, CSRanking).
- Authors: Zizhao Mo, Huanle Xu, Cheng-Zhong Xu
Optimal Resource Efficiency with Fairness in Heterogeneous GPU Clusters
- Accepted by Middleware’ 24 (CCF-B).
- Authors: Zizhao Mo, Huanle Xu, Wing Cheong Lau
Derm: SLA-aware Resource Management for Highly Dynamic Microservices
- Accepted by ISCA’ 24 (CCF-A, CSRanking).
- Authors: Liao Chen, Shutian Luo, Chenyu Lin, Zizhao Mo, Huanle Xu, Kejiang Ye, Cheng-Zhong Xu
Interference-aware Multiplexing for Deep Learning in GPU Clusters A Middleware Approach
- Accepted by SC’ 23 (CCF-A, CSRanking).
- Authors: Wenyan Chen, Zizhao Mo, Huanle Xu, Kejiang Ye, Cheng-Zhong Xu
🎖 Honors and Awards
- 2021 - 2025: Ph.D. Scholarship. University of Macau.
- 2024: Travel grant for ASPLOS’24.
- 2021: Arthur and Louis May Scholarship. Hong Kong University of Science and Technology.
🔖 Academic Activities
The reviewers of:
- IEEE Transactions on Computers (IEEE TC)
- Transactions on Architecture and Code Optimization (ACM TACO)
- Transactions on Services Computing (IEEE TSC)
- Journal of Systems Architecture (JSA)
- IEEE Transactions on Consumer Electronics
The program committee member of:
- 2027: HPCA
- 2026: IEEE/ACM Micro; MLsys (Artifact evaluation); IJCAI-ECAI (AI4Tech workshop)
📖 Educations
- 2021 - 2025: Ph.D. in Computer Science, University of Macau.
- 2020 - 2021: MSc in Information Technology, Hong Kong University of Science and Technology.
- 2014 - 2018: B.Eng. in Software Engineering, South China University of Technology.