Jiabo Shi
Email: j.shi.1@research.gla.ac.uk
Office: Room F101, School of Computing Science, Sir Alywn Williams Building, University of Glasgow, Glasgow, G12 8QQ
Research title: A Deep-Learning Tailored Scheduler for Out-Of-Memory Awareness in Shared Clusters
Research Summary
My current research interest lies in exploring and developing efficient and reliable resource scheduling systems for large-scale deep learning clusters.
My recent research focuses on addressing critical challenges, including the estimation of GPU memory and the frequent Out-of-Memory (OOM) errors that occur during deep learning training jobs, thereby exacerbating the global GPU shortage issue in GPU clusters.
Publications
2025
Shi, Jiabo, Pezaros, Dimitrios 
ORCID: https://orcid.org/0000-0003-0939-378X and Elkhatib, Yehia ORCID: https://orcid.org/0000-0003-4639-436X
(2025)
xMem: A CPU-Based Approach for Accurate Estimation of GPU Memory in Deep Learning Training Workloads.
In: 26th ACM/IFIP International Middleware Conference, Nashville, TN, USA, 15-19 December 2025,
(Accepted for Publication)
Conference Proceedings
Shi, Jiabo, Pezaros, Dimitrios ORCID: https://orcid.org/0000-0003-0939-378X and Elkhatib, Yehia ORCID: https://orcid.org/0000-0003-4639-436X
(2025)
xMem: A CPU-Based Approach for Accurate Estimation of GPU Memory in Deep Learning Training Workloads.
In: 26th ACM/IFIP International Middleware Conference, Nashville, TN, USA, 15-19 December 2025,
(Accepted for Publication)
Supervisors
Grants
Conferences
Teaching
Additional Information
