CS Seminar Series: Pengfei Li

About this event

Scaling Intelligence Under Resource Constraints: Measurement and Optimization for Sustainable Computing

Abstract:

The rapid growth of artificial intelligence (AI) is driving the expansion of large-scale data centers and increasing demand for electricity. Supplying this computing infrastructure incurs some environmental and public health costs, including carbon emissions, water consumption, and air pollution. These impacts vary across locations, raising questions about both the scale of AI’s resource demands and how surrounding communities share the burden.

This talk examines these challenges with a particular focus on water. I will discuss how to quantify the water footprint of AI computing, accounting for both onsite cooling and offsite electricity generation. Beyond total consumption, I will examine how seasonal peaks in water demand create infrastructure needs. Site-level estimates and national projections illustrate why relatively small annual water use can coexist with substantial demands on local supply capacity. Next, I will explain the public health burden of data center air pollution, showing why carbon footprints alone cannot capture local impacts.

Finally, I will discuss approaches to mitigating these burdens through intelligent workload management, including health-informed computing and equitable geographical load balancing. These approaches highlight opportunities to reduce AI’s environmental impacts while accounting for the communities that bear them.

Bio:

Pengfei Li is an Assistant Professor in the School of Information at the Rochester Institute of Technology. He received his Ph.D. in Computer Science from the University of California, Riverside and his M.S.E. in Robotics from Johns Hopkins University. He is a recipient of the ACM SIGMETRICS Doctoral Dissertation Award (honorable mention), recognizing his doctoral work in performance modeling and computer systems. His research focuses on trustworthy and sustainable AI, with an emphasis on learning-augmented online optimization, online resource allocation, and responsible machine learning frameworks for decision-making systems.