Yu Cheng

prof_pic.jpg

Hi, this is Yu Cheng (程羽). I am a fourth-year graduate student in School of Computer Science at Peking University, advised by Prof. Zhi Yang. I received my B.S. (Summa Cum Laude) from Turing Class at Peking University in 2022.

I am also a research intern in System Research Group of Microsoft Research Asia (MSRA), supervised by Dr. Jilong Xue and Dr. Lingxiao Ma.

I am a core developer of TileLang, a programming language for AI workload on various hardware platforms.

My research interests lie in deep learning systems, specifically on compilation optimization of deep learning framework.

Email: yu [DOT] cheng [AT] pku [DOT] edu [DOT] cn

selected publications

  1. OSDI
    PipeThreader: Software-Defined Pipelining for Efficient DNN Execution
    Yu Cheng, Lei Wang, Yining Shi, and 9 more authors
    The 19th USENIX Symposium on Operating Systems Design and Implementation (OSDI ’25), 2025
  2. ICLR
    TileLang: Bridge Programmability and Performance in Modern Neural Kernels
    Lei Wang, Yu Cheng, Yining Shi, and 9 more authors
    The Fourteenth International Conference on Learning Representations (ICLR 2026), 2026
    Oral Presentation
  3. Arxiv
    HySparse: A Hybrid Sparse Attention Architecture with Oracle Token Selection and KV Cache Sharing
    Yizhao Gao, Jianyu Wei, Qihao Zhang, and 11 more authors
    arXiv preprint arXiv:2602.03560, 2025
  4. SOSP
    Scaling Deep Learning Computation over the Inter-Core Connected Intelligence Processor
    Yiqi Liu, Yuqi Xue, Yu Cheng, and 4 more authors
    30th ACM Symposium on Operating Systems Principles (SOSP 2024), 2024

projects

  1. TileScale
    A high-performance distributed communication library for AI workloads.
  2. TileLang
    A domain-specific language designed to streamline the development of high-performance kernels for AI workloads.

experiences

04/2022-Now Microsoft Research Asia
07/2021-01/2022 Alibaba Cloud
10/2020-07/2021 Alibaba
  • Research intern
  • Advised by Liang Wang