About

Hi! I’m Chuanhao Sun, a Postdoc Researcher at the University of Edinburgh, School of Informatics (ICSA), where I received my PhD degree (July 2024).

My research bridges both network systems and machine learning. In network systems, my focus is primarily on edge computing, mobile networks, and ISP networks. In machine learning, I tackle cutting-edge challenges in areas such as time series analysis, image processing, and 3D rendering. My work has been published at leading conferences in both domains, including ACM MobiCom and ACM CoNEXT for network systems, and ICML for machine learning. I’m passionate about using machine learning to address system-level challenges (AI4System) and to optimizing systems for enhanced AI training and inference (System4AI).

My research

My reseaarch mainly includes two threads: (1) AI + System; (2) Machine learning.

AI + System

This part is mainly about using AI to resolve problem in systems. For example, we employ tailored generative models for system diagnostics, where advanced anomaly detection techniques for multivariate time series are crucial in addressing the growing complexity of modern systems and network traffic. Recent efforts also include applying the concept of large models, such as Mixture of Experts (MoE), to novel scenarios like highly efficient image processing.

Additionally, I am exploring the co-design of machine learning algorithms and systems, such as optimizing GPU caching schemes to improve the efficiency of 3D neural rendering.

Machine learning

The machine learning thread of my research focuses on addressing core algorithmic challenges independent of specific system contexts. Key topics include, but are not limited to, anomaly detection in multivariate time series, 3D rendering (e.g., NeRF and Gaussian Splats), and general processing tasks for time series and images, such as compression and super-resolution.

My background and history

I received my Ph.D from the UoE School of Informatics after 4 years of study (Oct. 2019 to Dec. 2023), under the supervision of Prof. Mahesh K. Marina (on Network System, homepage), co-supervised by Dr. Kai Xu (on Machine Learning, homepage). After completing my Ph.D., I transitioned directly into a Postdoctoral Research Associate position within the same research group.

During my Ph.D. and subsequent postdoc, I developed extensive collaborations across both academia and industry. In academia, I co-authored with researchers worldwide, including those from UIUC, UCL, Polimi, and IMDEA, on a range of topics spanning systems, communication, and machine learning theory. In industry, I worked as a research intern and long-term collaborator with Microsoft Research, under the mentorship of Dr. Bozidar Radunovic (homepage) and Dr. Xenofon Foukas (homepage), where we conducted impactful system research, with our latest work set to appear at MobiCom ‘24. Prior to my work with Microsoft, I collaborated with Samsung Research through a sponsorship, resulting in a series of publications in top conferences and journals on the application of generative models in mobile networks.

Before join the University of Edinburgh at 2019, I spent my early years in Beijing University of Posts and Telecommunications (BUPT), where I start to develop my research experience under the supervsion of Prof. Wenbo Wang (homepage) and Prof. Xing Zhang (homepage). In the early years of my research at BUPT, my work focused on optimal resource scheduling in edge networks, as well as reinforcement learning-based streaming solutions. During this time, I also worked part-time as a research intern at Intel Lab China, where my primary focus was on physical layer simulations for radio access networks.