Inside Our ProgramProgram Events
Events
-
Feb3
EVENT DETAILSmore info
lessTitle: Privacy-Preserving Federated Learning at Scale: From Algorithms to Supercomputers
Bio: I am a Computational Mathematician in Laboratory for Applied Mathematics, Numerical Software, and Statistics, Mathematics and Computer Science Division at Argonne National Laboratory, and a Senior Scientist at-Large at the University of Chicago Consortium for Advanced Science and Engineering. My research focuses on federated learning algorithms and software development, as well as modeling and numerical algorithms for large-scale optimization on high-performance computing systems and GPUs. My work is applied to areas including electric grid systems, healthcare, and key scientific domains of interest to the Department of Energy. Before joining Argonne, I obtained a Ph.D. degree in Industrial Engineering and Management Sciences from Northwestern University. I am a recipient of DOE Early Career Research Program award. I serve as associate editors in Mathematical Programming Computation and Naval Research Logistics and board members for COIN-OR Foundation and IISE Energy Systems.
Abstract: Federated learning (FL) is increasingly attractive for AI for Science because it enables collaborative model training across institutions and facilities without centralizing sensitive data. In this talk, I will give a brief overview of our DOE-funded AI for Science effort and the multidisciplinary team building scalable learning workflows for scientific and energy applications. I will then introduce privacy-preserving federated learning, focusing on practical mechanisms (e.g., secure aggregation and differential privacy) and the core tradeoffs they impose on accuracy, communication, and systems performance.
The second half of the talk will focus on our recent queue-aware FL protocol designed for cross-facility training on leadership-class HPC systems. Unlike conventional synchronous or fully asynchronous FL, our algorithm treats batch-scheduler delays as a first-class systems signal: it adapts local work and aggregation to time-varying queue and execution delays to reduce straggler effects while controlling staleness. I will present the key ideas and empirical results from deploying FL across Aurora, Frontier, Polaris, and Perlmutter, highlighting end-to-end performance, stability under heterogeneous runtimes, and what this implies for building reliable, privacy-preserving, multi-site AI training pipelines for science.
TIME Tuesday, February 3, 2026 at 11:00 AM - 12:00 PM
LOCATION A230, Technological Institute map it
CONTACT Kendall Minta kendall.minta@gmail.com EMAIL
CALENDAR Department of Industrial Engineering and Management Sciences (IEMS)
-
Feb5
EVENT DETAILS
lessCan AI understand your listening soul? We will explore how genAI capabilities translate into real-world, industry-scale use case from a major audio streaming platform. By introducing a novel semantic tokenization that encodes relationships between content and users, data scientists and AI practitioners can enable LLMs to reason directly over music catalogs, user intent, and personalization signals. This unlocks explainable recommendations, richer semantic search, and more adaptive discovery experiences, built at production scale.
Learn how generative and agentic AI will power the next generation of interactive, personalized, and creative entertainment experiences.
Speaker Biography:Kingsley Di is the Head of Applied AI/ML at SiriusXM & Pandora, and a veteran machine learning engineer. In his current role, he is leading SiriusXM’s end-to-end AI strategy and overseeing large-scale model development, product innovation and global enterprise deployment. Prior to joining SiriusXM, Kingsley led an NLP research team at S&P Global where he helped modernize the credit ratings value chain through intelligent automation. Earlier in his career, he joined American Express as a Senior Research Scientist after earning his Ph.D. from Northwestern University. His research was focused on distributed machine learning and deep learning with concentration in transportation and sports entertainment.
Aside from AI and ML models, Kingsley is an active photographer and has a passion for documenting and reflecting AI’s evolving impact on civilization through his photos.
TIME Thursday, February 5, 2026 at 2:30 PM - 3:30 PM
LOCATION Krebs Classroom, North Campus Parking Garage map it
CONTACT Master of Science in Machine Learning and Data Science Program mlds@northwestern.edu EMAIL
CALENDAR Master of Science in Machine Learning and Data Science (MLDS)
-
Feb10
TIME Tuesday, February 10, 2026 at 11:00 AM - 12:00 PM
LOCATION A230, Technological Institute map it
CONTACT Kendall Minta kendall.minta@gmail.com EMAIL
CALENDAR Department of Industrial Engineering and Management Sciences (IEMS)
-
Mar3
TIME Tuesday, March 3, 2026 at 11:00 AM - 12:00 PM
LOCATION A230, Technological Institute map it
CONTACT Kendall Minta kendall.minta@gmail.com EMAIL
CALENDAR Department of Industrial Engineering and Management Sciences (IEMS)
-
Mar10
TIME Tuesday, March 10, 2026 at 11:00 AM - 12:00 PM
LOCATION A230, Technological Institute map it
CONTACT Kendall Minta kendall.minta@gmail.com EMAIL
CALENDAR Department of Industrial Engineering and Management Sciences (IEMS)
