Shaoxun Zeng | WeLM, Tencent

About Me

I am currently working in the WeLM team at Tencent, where I am advancing systems for LLM training, inference, and reinforcement learning. I received my Ph.D. from Tsinghua University in 2026, supervised by Prof. Youyou Lu and Prof. Jiwu Shu.

My research focuses on system optimizations for LLMs, particularly on infrastructure scaling, resource efficiency, and full-stack co-design. If you are interested in democratizing access to state-of-the-art AI for billions of users and conducting cutting-edge research, join us.

Education

Education:

Ph.D. in Computer Science, Tsinghua University, 2021–2026
Summer School, University of California, Berkeley, Summer 2019
B.S. in Software Engineering, Nanjing University, 2017–2021

Publications

2026

[FAST'26]GPU Checkpoint/Restore Made Fast and LightweightShaoxun Zeng, Tingxu Ren, Jiwu Shu, Youyou Lu
The 24th USENIX Conference on File and Storage Technologies, 2026 (Distinguished Artifact Award)
[ACM ToS'26]Speeding Up Write-Read Orders in Disaggregated StorageShaoxun Zeng, Xiaojian Liao, Hao Guo, Jiwu Shu, Youyou Lu
ACM Transactions on Storage, 2026
[ICML'26]WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast InferenceAiwei Liu, Minghua He, Shaoxun Zeng, Linhao Zhang, Chuhan Wu, Wei Jia, Yuan Liu, Yang Yu, Xiao Zhou, Jie Zhou
International Conference on Machine Learning, 2026 (Oral)

2025

[ASPLOS'25]Medusa: Accelerating Serverless LLM Inference with MaterializationShaoxun Zeng, Minhui Xie, Shiwei Gao, Youmin Chen, Youyou Lu
The 30th Conference on Architectural Support for Programming Languages and Operating Systems, 2025
[USENIX ATC'25]Weaver: Efficient Multi-LLM Serving with Attention OffloadingShiwei Gao, Qing Wang, Shaoxun Zeng, Youyou Lu, Jiwu Shu
USENIX Annual Technical Conference, 2025
[ASPLOS'25]Frugal: Efficient and Economic Embedding Model Training with Commodity GPUsMinhui Xie, Shaoxun Zeng, Hao Guo, Shiwei Gao, Youyou Lu
The 30th Conference on Architectural Support for Programming Languages and Operating Systems, 2025

2024

[EuroSys'24]Volley: Accelerating Write-Read Orders in Disaggregated StorageShaoxun Zeng, Xiaojian Liao, Hao Guo, Youyou Lu
The 19th European Conference on Computer Systems, 2024
[CCCF]面向高速硬件的高性能文件系统Youyou Lu, Hao Guo, Shaoxun Zeng, Yitian Yang
中国计算机学会通讯, 2024

2023

[USENIX ATC'23]SingularFS: A Billion-Scale Distributed File System Using a Single Metadata ServerHao Guo, Youyou Lu, Wenhao Lv, Xiaojian Liao, Shaoxun Zeng, Jiwu Shu
USENIX Annual Technical Conference, 2023

Honors Awarded

FAST'26 Distinguished Artifact Award, 2026.
No.1 in the ASPLOS / EuroSys 2025 Contest of LLM inference optimization, 2025.
Student Travel Grant for ASPLOS'25/EuroSys'25, 2025.
No.1 in the IO500 Research Overall Score (SuperFS), ISC'23/SC'23/ISC'24/SC'24/ISC'25/SC'25, 2023-2025.