Publications

* denotes equal contribution

2027

  1. magneton.png
    Magneton: Optimizing Energy Efficiency of ML Systems via Differential Energy Debugging
    Yi Pan*, Wenbo Qian*, Dedong Xie, Ruiyan Hu, Yigong Hu, and Baris Kasikci
    In 24th USENIX Symposium on Networked Systems Design and Implementation (NSDI’27) , 2027
    arXiv

2026

  1. TraceLab: Characterizing Coding Agent Workloads for LLM Serving
    Kan Zhu, Mathew Jacob, Chenxi Ma, Yi Pan, Stephanie Wang, Arvind Krishnamurthy, and Baris Kasikci
    arXiv preprint, 2026
    arXiv
  2. Piper: A Programmable Distributed Training System
    Megan Frisella, Shubham Tiwari, Andy Ruan, Yi Pan, Parker Gustafson, Mat Jacob, Gilbert Bernstein, and Stephanie Wang
    arXiv preprint, 2026
    arXiv
  3. dynaflow.png
    DynaFlow: Transparent and Flexible Intra-Device Parallelism via Programmable Operator Scheduling
    Yi Pan, Yile Gu, Jinbin Luo, Yibo Wu, Ziren Wang, Hongtao Zhang, Ziyi Xu, Shengkai Lin, Baris Kasikci, and Stephanie Wang
    In 9th Annual Conference on Machine Learning and Systems (MLSys’26) , 2026
    arXiv
  4. muxtune.png
    MuxTune: Efficient Multi-Task LLM Fine-Tuning via Spatial-Temporal Backbone Multiplexing
    Chunyu Xue*, Yi Pan*, Weihao Cui, Quan Chen, Shulai Zhang, Bingsheng He, and Minyi Guo
    In 23nd USENIX Symposium on Networked Systems Design and Implementation (NSDI’26) , 2026
    arXiv

2025

  1. SHIFT: Exploring the Boundary of RDMA Network Fault Tolerance
    Shengkai Lin, Kairui Zhou, Hongtao Zhang, Yibo Wu, Yi Pan, Yihan Yang, Qinwei Yang, Wei Zhang, Arvind Krishnamurthy, and Shizhen Zhao
    arXiv preprint, 2025
    arXiv