Fan Lai earns Towner Prize for Outstanding PhD Research

The award recognizes creative and outstanding research achievements.
Fan Lai

Fan Lai, a PhD candidate in computer science and engineering, has been awarded the Richard F. and Eleanor A. Towner Prize for Outstanding PhD Research by the College of Engineering for his creativity and outstanding research achievements. Lai works to enable fast, distributed computing over networks, especially for cloud computing and machine learning tasks. His research has been adopted widely in the open-source community and at companies including Meta and LinkedIn.

Lai’s work spans different layers of the software stack needed for efficient datacenter-scale and planet-scale computing. To date, he’s devised a number of deployable systems to tackle the issues that come with running big data and machine learning (ML) programs in highly distributed settings. In data centers, this often means accessing computing and memory resources spread across multiple machines. More recently, he’s also focused on systems to run ML models across the planet-scale devices connected to the Internet.

The latter problem, called federated learning, involves distributing the training of ML models to thousands or millions of small devices rather than centralizing user data. FedScale, a project led by Lai and his advisor Prof. Mosharaf Chowdhury, gave research on federated learning a big boost by building the system infrastructure to make deploying federated learning across laptops and smartphones practical, as well as a gym to simulate training environment in the cloud.

The platform can simulate the behavior of millions of user devices on a few GPUs and CPUs, enabling developers of these models to explore how their federated learning program will perform without the need for large-scale deployment. It serves a variety of real-world ML applications, including image classification, object detection, language modeling, speech recognition and machine translation.

The paper presenting FedScale, “FedScale: Benchmarking Model and System Performance of Federated Learning at Scale,” earned a Best Paper Award at ResilientFL 2022. Lai has built other systems to improve the federated learning workflow, including Oort, which makes the selection of participant devices more efficient by predicting the most useful data and hardware available in the pool of candidates.

In addition to a Best Paper Award at ResilientFL, Lai has also received a Distinguished Artifact Award  at OSDI 2021, and was a finalist for a Meta PhD Research Fellowship.