Ligeng Zhu
ligeng (dot) zhu [at] gmail.com (dot) com
(add [plus] notAI to bypass spam filter)

I am a Researcher at NVIDIA, fortunately working with Prof. Song Han. Before NVIDIA, I was at MIT (freezing Boston), Simon Fraser University (warmful Vancouver), and Zhejiang University (beautiful Hangzhou).

My research current interests focus on agentic, efficient, and multi-modal LLMs. During my studies, I worked with Prof. Song Han on efficient machine learning, Prof. Brian Funt on colour vision, and Prof. Ping Tan on attribute recognition.

If you find any research interests that we might share, feel free to drop me an email. I am always open to potential collaborations.

Email  /  Google Scholar  /  GitHub  /  Résumé
Latest update on Jan 20 2026.

profile photo
News
Publications

PockEngine: Sparse and Efficient Fine-tuning in a Pocket
Ligeng Zhu, Lanxiang Hu, Ji Lin, Wei-Chen Wang, Wei-Ming Chen, Chuang Gan, Song Han
56th IEEE/ACM International Symposium on Microarchitecture (MICRO-56), 2023
Paper  /  Website  /  Slides  /  YouTube
  • Media coverage: MIT News  /  Zhihu

    • On-Device Training Under 256KB Memory
    Ji Lin* , Ligeng Zhu* , Wei-Ming Chen , Wei-Chen Wang , Chuang Gan, Song Han
    (* denotes equal contribution, sorted in aplhabetic order)
    Annual Conference on Neural Information Processing Systems (NeurIPS), 2022
    Paper  /  Website  /  Slides  /  YouTube  /  GitHub
  • Media coverage: MIT Homepage  /  Zhihu

    • Enable deep learning on mobile devices: Methods, systems, and applications
    Han Cai*, Ji Lin*, Yujun Lin*, Zhijian Liu*, Haotian Tang*, Hanrui Wang*, Ligeng Zhu*, Song Han
    (* denotes equal contribution, sorted in aplhabetic order)
    ACM Transactions on Design Automation of Electronic Systems (TODAES), 2022
    Paper

    Delayed Gradient Averaging: Tolerate the Communication Latency for Federated Learning
    Ligeng Zhu, Hongzhou Lin, Yao Lu, Yujun Lin, Song Han
    Annual Conference on Neural Information Processing Systems (NeurIPS), 2021
    Paper  /  Webpage

    IOS: Inter-Operator Scheduler for CNN Acceleration
    Yaoyao Ding, Ligeng Zhu, Zhihao Jia, Gennady Pekhimenko, Song Han
    Conference on Machine Learning and Systems (MLSys), 2021
    Paper  /  Webpage  /  Code  /  YouTube  /  BiliBili

    TinyTL: Reduce Memory, Not Parameters for Efficient On-Device Learning
    Han Cai, Chuang Gan, Ligeng Zhu, Song Han
    Advances in Neural Information Processing Systems (NeurIPS), 2020
    Paper

    DataMix: Efficient Privacy-Preserving Edge-Cloud Inference
    Zhijian Liu, Zhanghao Wu, Chuang Gan, Ligeng Zhu, Song Han:
    European Conference on Computer Vision (ECCV) , 2020
    Paper

    HAT: Hardware-Aware Transformers for Efficient Neural Machine Translation
    Hanrui Wang*, Zhanghao Wu*, Zhijian Liu*, Han Cai, Ligeng Zhu, and Song Han
    Annual Conference of the Association for Computational Linguistics (ACL), 2020
    Paper  /  Webpage  /  Code  /  YouTube  /  BiliBili

    Distributed Training across the World
    Ligeng Zhu, Yao Lu, Yujun Lin, Song Han
    Neural Information Processing Systems (NeurIPS) Workshop on Systems for ML (MLSys), 2019
    Paper  /  Poster
  • Scale Synchronous SGD across the world, without loss of speed and accuracy!

    • Deep Leakage from Gradients
    Ligeng Zhu, Zhijian Liu, Song Han
    Neural Information Processing Systems (NeurIPS), 2019
    arXiv  /  Poster  /  Code (implementation in just 20 lines)
  • Media coverage: Zhihu

    • AutoML for Architecting Efficient and Specialized Neural Networks
    Han Cai*, Ji Lin*, Zhijian Liu*, Yujun Lin*, Kuan Wang*, Tianzhe Wang*, Ligeng Zhu*, Song Han
    (* denotes equal contribution, sorted in aplhabetic order)
    IEEE International Symposium on Microarchitecture (Micro), 2019
    arXiv  /  Paper  /  Slides

    ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware.
    Han Cai, Ligeng Zhu, Song Han
    International Conference on Learning Representations (ICLR) , 2019 (378 citations)
    arXiv  /  Project Page  /  GitHub (1.1k stars) (integrated into PyTorch Hub)  /  Poster
  • Media coverage: MIT News, IEEE Spectrum, Zhihu, 机器之心

    • Sparsely Aggregated Convolutional Networks
    Ligeng Zhu, Ruizhi Deng, Michael Maire, Zhiwei Deng, Greg Mori, Ping Tan
    European Conference on Computer Vision (ECCV), 2018
    arXiv  /  Code  /  Poster

    Small Object Sensitive Segmentation of Urban Street Scene With Spatial Adjacency Between Object Classes.
    Dazhou Guo*, Ligeng Zhu*, Yuhang Lu, Hongkai Yu, Song Wang
    IEEE Transactions on Image Processing (TIP), 2018
    Paper

    Does Colour Really Matter? Evaluation via Object Classification.
    Brian Funt, Ligeng Zhu
    Color and Imaging Conference (CIC), 2018
    Paper  /  Poster

    Colorizing Color Images.
    Ligeng Zhu, Brian Funt
    Human Vision and Electronic Imaging (HEVI), 2018
    Paper  /  Code  /  Poster

    Attribute Recognition from Adaptive Parts.
    Luwei Yang, Ligeng Zhu, Yichen Wei, Shuang Liang, Ping Tan.
    British Machine Vision Conference (BMVC), 2016
    arXiv  /  Poster

    Students Collaborated with
    Talks & Presentations
    • [08/2019] AutoML for Efficient Neural Architecture Design (Slides)
        @ OpenPower Summit, Polarr Tech
    • [08/2019] Scalable and Secure Machine Learning for Edge Devices @ Qualcomm
    • [05/2019] Neural Architecture Designs @ UIUC IFP Group (Slides)
    • [12/2018] Proxylessly Specialize CNN for Hardware @ IBM-MIT Watson Events (Poster)
    • [01/2018] Sparsely Aggregated Convolutional Networks (Slides)
        @ UBC Vision Group, Deephi Tech, Sensetime Inc
    • [11/2017] Invited lectures about deep learning (Lecture1, Lecture2)
        @ SFU Computer Vision Course (CMPT-412), ZJU Programming Group
    Open-source Projects (Selected)

    My life (both academic and daily) is greatly powered by open source projects. To thank their selfless effort, I embrace open source as much as possible. Please refer to my github for a complete list of projects.

    PyTorch-OpCounter
    Star Fork
    Count the MACs / FLOPs of your PyTorch model.
    Avaliable through PyPi: pip install thop
    ProxylessNAS
    Star Fork
    Directly and efficiently search neural network architectures.
    Integrated into PyTorch Hubs!
    PyTorch-Memonger
    Star Fork
    Sublinear memory optimization for deep learning.
    Efficient-PyTorch
    Star Fork
    My best practice of training large dataset using PyTorch.
    Services

    Review papers for:

  • 2025: CVPR, ICLR, ICCV
  • 2024: CVPR, ICLR, ICML
  • 2023: NeurIPS, CVPR, ICLR, ICML
  • 2022: NeurIPS, CVPR
  • 2021: NeurIPS, ICCV, ICML, ACL
  • 2020: NeurIPS, CVPR, AAAI
  • 2019: NeurIPS, ICCV, CVPR
  • Journals: T-PAMI, IEEE Micro


    • Design and source code from Jon Barron. Style adapted from Zhijian Liu. Vibing and polishing w/ Antigravity.