Welcome to Zhouzi’s Homepage
I am Zhouzi Li (李洲子 in Chinese, how to pronounce), and I also go by Arthur. I am a fourth-year PhD student in the Computer Science Department at Carnegie Mellon University, advised by Prof. Mor Harchol-Balter. My research focuses on queueing theory and computer system performance, with a particular interest in scheduling problems arising in real-world systems — including cloud GPU resource allocation and LLM inference jobs scheduling.
Before CMU, I received my B.Eng. from Yao Class (IIIS), Tsinghua University.
Recent News
- Sep 2026 Our paper SPLIT: SymPathy for Large jobs Improves Tail latency has been accepted by SIGMETRICS 2027!
- Aug 2026 Our paper Mean Field Optimal Core Allocation across Malleable Jobs has been accepted by IFIP Performance 2026!
- May-Aug 2026 Internship at Google in Seattle, on TPU cluster scheduling.
- Apr 2026 Our paper Improving Upon the Generalized c-μ Rule: a Whittle Approach has been accepted by Operations Research!
- Feb 2026 Our paper BOA Constrictor: Squeezing Performance out of GPUs in the Cloud via Budget-Optimal Allocation is now on arXiv.
Research Overview
Tail optimization in multi-server queues
Scheduling policies that minimize the response time tail in multi-server systems with heavy-tailed job sizes. SIGMETRICS 2027
Resource allocation for ML training jobs
- Theory: Optimal core allocation across malleable jobs, which can run on any number of cores. IFIP Performance 2026
- Systems: Budget-optimal allocation of rented cloud GPUs for ML training. Under submission
LLM inference scheduling
Scheduling and batching LLM inference requests to meet latency targets while using GPUs efficiently. Under submission
Bandit theory for scheduling with time-varying holding costs
Scheduling jobs whose cost of waiting grows as they wait, using a restless bandit (Whittle index) approach. Operations Research 2026 Queueing Systems 2026
My CV
You can download my CV here.
Outside of research, I enjoy playing basketball and tennis. I am also a great fan of music and play the guitar.