School of CSE Seminar Series: I-Hsin Chung

Speaker: I-Hsin Chung, staff researcher at IBM Research
Date and Time: April 11, 2:00-3:00 p.m.
Location: TBD
Host: Ümit Çatalyürek

Title: Application-Driven High Performance AI system Design

Abstract: In this talk, we explore the challenges of future system design, focusing on integrating application performance characteristics with emerging technologies to achieve a balanced, high-performance architecture. Through practical examples, we demonstrate how application-driven design principles can be applied to both system software and hardware. Specifically, we present the design of IBM Vela, a cloud-based HPC system tailored for machine learning and AI workloads. By analyzing the workflow and identifying co-design opportunities across multiple system layers, we enable the virtualized system to achieve near bare-metal performance. The talk will highlight key performance insights and conclude with a discussion on potential research directions for advancing system design in the era of AI and machine learning.

Bio: Dr. I-Hsin Chung is a computer scientist with a Ph.D. in Computer Science from the University of Maryland, College Park. Since joining IBM Research, his work has focused on critical areas such as performance modeling, tuning, and advanced tools within computer science. His primary research expertise lies in system architecture, with a particular emphasis on cloud computing and high-performance computing. Currently, he is leading efforts to co-design future cloud computer infrastructures tailored for strategic application workloads. Dr. Chung’s contributions extend across various domains of system architecture, including system software development, performance analysis, tool development, and modeling, with a focus on IBM platforms like Vela, CORAL, POWER, and Blue Gene systems. Additionally, he serves as an adjunct professor at the Courant Institute at NYU and Columbia University, where he shares his knowledge and expertise beyond his work at IBM Research.