Luo Mai

Assistant Professor

University of Edinburgh

About Me

I am an Assistant Professor (Lecturer) in the School of Informatics at the University of Edinburgh. I am a member of the Institute of Computing Systems Architecture where I am leading the Large-Scale System Software Group.

In addition to my academic role, I co-found the Open Machine Learning System Community, serving as an educational platform that fosters the development of system software for machine learning. I have also contributed to the education of machine learning systems through my open-source textbook, Machine Learning Systems: Design and Implementation.

Before coming to Edinburgh, I was a research associate (2018 - 2020) at the Imperial College London working with Peter Pietzuch. I obtained my PhD from Imperial College London under the supervision of Paolo Costa and Alexander L. Wolf. My PhD was supported by a Google Fellowship in Cloud Computing. During my PhD study, I was a research intern (2015, 2016) and a visiting researcher (2017) at Microsoft Research.

05/2023: I have Ph.D. and Postdoc openings to work on cloud data systems, AI systems and quantum system software. Contact me if you are interested.

Interests

Computer Systems
Machine Learning
Data Management

Education

PhD in Computer Science, 2018

Imperial College London, UK
MRes in Advanced Computing, 2012

Imperial College London, UK
BSc in Software Engineering, 2011

Xidian University, China

News

TorchOpt and Gear to appear in NeurIPS and ICML

We have three papers to appear in leading machine learning conferences: “GEAR: A GPU-Centric Experience Replay System for Large Reinforcement Learning Models” to appear in ICML 2023. “A Theoretical Understanding of Gradient Bias in Meta-Reinforcement Learning” appeared in NeurIPS 2022.

Last updated on May 10, 2023

MegBA in ECCV 2022

Our Paper “MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment” is accepted by ECCV 2022.

Jul 10, 2022

Ekko in OSDI 2022

Our Paper “Ekko: A Large-Scale Deep Learning Recommender System with Low-Latency Model Update” is accepted by USENIX Symposium on Operating Systems Design and Implementation (OSDI) 2022. OSDI brings together professionals from academic and industrial backgrounds in what has become a premier forum for discussing the design, implementation, and implications of systems software.

Mar 18, 2022

See all posts

Publications

Hanjing Wang*, Man-Kit Sit*, Congjie He, Ying Wen, Weinan Zhang, Jun Wang, Yaodong Yang, Luo Mai (2023). GEAR: A GPU-Centric Experience Replay System for Large Reinforcement Learning Models. In ICML.

Chijun Sima*, Yao Fu*, Man-Kit Sit, Liyi Guo, Xuri Gong, Feng Lin, Junyu Wu, Yongsheng Li, Haidong Rong, Pierre-Louis Aublin, Luo Mai (2022). Ekko: A Large-Scale Deep Learning Recommender System with Low-Latency Model Update. In USENIX OSDI.

PDF

Bo Liu, Xidong Feng, Jie Ren, Luo Mai, Rui Zhu, Haifeng Zhang, Jun Wang, Yaodong Yang (2022). A Theoretical Understanding of Gradient Bias in Meta-Reinforcement Learning. In NeurIPS.

PDF

Jie Ren, Xidong Feng, Bo Liu, Xuehai Pan, Yao Fu, Luo Mai, Yaodong Yang (2022). TorchOpt: An Efficient Library for Differentiable Optimization. In NeurIPS OPT Workshop.

PDF Code

Jie Ren*, Wenteng Liang*, Ran Yan, Luo Mai, Shiwen Liu, Xiao Liu (2022). MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment. In ECCV.

PDF Code

Le Xu, Shivaram Venkataraman, Indranil Gupta, Luo Mai, Rahul Potharaju (2021). Move Fast and Meet Deadlines: Fine-grained Real-time Stream Processing with Cameo. In USENIX NSDI.

PDF

Zihan Ding, Tianyang Yu, Yanhua Huang, Hongming Zhang, Guo Li, Quancheng Guo, Luo Mai, Hao Dong (2021). Efficient Reinforcement Learning Development with RLzoo. In ACM Multimedia (Open-source Software Competition).

PDF Code

Yixiao Guo, Jiawei Liu, Guo Li, Luo Mai, Hao Dong (2021). Fast and Flexible Human Pose Estimation with HyperPose. In ACM Multimedia (Open-source Software Competition).

PDF Code

Luo Mai, Guo Li, Marcel Wagenlander, Konstantinos Fertakis, Andrei-Octavian Brabete, Peter Pietzuch (2020). KungFu: Making Training in Distributed Machine Learning Adaptive. In USENIX OSDI.

PDF Code