Mohammad Hajiesmaili

Abstract Title

Algorithmic Foundations of Grid-Integrated AI Infrastructure
The generative AI era is redefining both the scale and urgency of computing. Edge–cloud systems that power today’s digital world are consuming energy at unprecedented rates, pushing power grids to their limits and revealing a growing interdependence between computing and energy infrastructures. To sustain progress, we must move beyond incremental optimizations toward a fundamental redesign that treats compute and energy as components of a unified ecosystem.
This talk introduces the new theoretical challenges that arise in realizing the vision of grid-integrated computing infrastructure. The central idea is to exploit the intrinsic flexibility of computational workloads and integrate them with distributed energy resources—such as renewables and storage—to enhance both grid stability and compute reliability. Building on this perspective, we present a theoretical framework grounded in learning-augmented and distributed algorithms, enabling systems that adapt dynamically to real-time grid and environmental signals. These algorithms provide the foundation for networked adaptive resource allocation, scheduling, and load balancing across energy-intensive yet flexible computing and power systems.  

Biography

Mohammad Hajiesmaili is an Associate Professor in the Manning College of Information and Computer Sciences at the University of Massachusetts Amherst, where he directs the Sustainability, Optimization, Learning, and Algorithms Research (SOLAR) Lab. His research develops optimization and machine learning methods to address challenges in real-world computing systems, with a particular focus on decarbonizing digital and societal infrastructure. His honors include leading the theory and AI thrust of an NSF Expeditions in Computing project on Computational Decarbonization, an NSF CAREER Award, and research awards from NSF, Amazon, Google, VMware, and Adobe. He has also received multiple Best Paper Runner-up awards at top-tier venues, including ACM SIGMETRICS, ACM e-Energy, and ACM BuildSys.

Contact

hajiesmaili@cs.umass.edu