SDI / PDL Talks - Olivia Hsu, Aaron Ogus
August 20, 2026 12:00PM—2:00PM
Location:
Virtual Presentations - ET
-
Remote Access - Zoom
Speaker:
OLIVIA HSU and AARON OGUS
Talk 1
Olivia Hsu, Assistant Professor, Department of Electrical and Computer Engineering, Carnegie Mellon University
— Programming Data-dependent Applications on Hardware Accelerators
Modern workloads, including machine learning and data analytics, are increasingly data-dependent. Their execution depends heavily on sparsity, dynamic control flow, irregular communication, and evolving runtime behavior. At the same time, computer architecture is rapidly shifting toward heterogeneous, domain-specific hardware, which include hardware accelerators, reconfigurable dataflow architectures, and decoupled access-execute systems. This talk discusses the systems challenges that emerge at the intersection of these trends and argues that hardware adoption is increasingly limited not by hardware capability itself, but by a systems ability to support it. I will present my group’s research on hardware–software systems for efficiently executing complex, data-dependent applications on hardware accelerators. The talk covers research directions in my group, including programming systems for sparse and dynamic ML accelerators and compiler/runtime abstractions for irregular computation on dataflow accelerators.
…
Olivia Hsu is an Assistant Professor of Electrical and Computer Engineering and, by courtesy, Computer Science at Carnegie Mellon University. Her research lies at the intersection of compilers and computer architecture, spanning programming languages to digital VLSI. Her recent work includes sparse tensor compilation, compilation and mapping for heterogeneous domain-specific accelerators, dataflow abstractions for hardware, and reconfigurable spatial architectures. Her research has been recognized with the 2026 ACM SIGARCH/IEEE CS TCCA Outstanding Dissertation Award, a Distinguished Paper Award at PLDI 2023, and a Best Paper at the Deep-Learning for Code Workshop co-located with ICML 2025. Her honors include the NSF Graduate Research Fellowship, Rising Stars in EECS, IEEE-HKN Alton B. Zerby and Carol M. Korner Outstanding Student Award, and a UC Berkeley Outstanding Graduate Student Instructor Award. Prior to joining Carnegie Mellon University, she was a postdoctoral researcher at Stanford University, where she also earned her Ph.D. and M.S. in Computer Science, advised by Kunle Olukotun and Fredrik Kjolstad.
Talk 2
Aaron Ogus, Distinguished Engineer working, Microsoft Azure Core Team
— Reliably Storing Zettabytes of Data: Moore's Law Meets Murhpy's Law (subject to change)
…
Aaron Ogus is a Distinguished Engineer working in the Microsoft Azure Core Team. He works (with many others) on optimizing the co-design of Hardware and Software for Azure Storage, Compute and Networking. Aaron has been working on Azure Storage since its inception in 2008 and has helped Microsoft optimize storage design and deployment, and has worked with industry partners to optimize components, servers and switch designs for the cloud. Aaron has helped influence the evolution of Network Switches, SSDs, HDDs and Archival Systems for Cloud deployment. Aaron started at Microsoft in 1991, and before Azure worked on a few notable projects, including Windows 95, DirectX, XBOX and Forza Motorsport.
Aaron has a regional SCCA racing championship in Formula Mazda, which was instrumental in helping make Forza the highest grossing racing franchise on XBOX. Aaron also plays a little poker from time to time. He finished second in a World Poker Tour Tournament at a Televised Final Table.
Zoom Participation. See announcement.
For More Information:
karenl@andrew.cmu.edu