Title: Liquid Machine: A 15-year Journey Towards Software-Defined Everything
Abstract: Computer Engineering is undergoing significant transformations, driven by escalating demands in artificial intelligence, cybersecurity, energy efficiency, sustainability, and fundamental shifts in technology scaling. For decades, the fundamental computing stack—the abstractions and interfaces that define our hardware and software systems—has remained largely unchanged. Traditionally, these abstractions have been “operational,” crafted to facilitate compilation, resource management, and backward compatibility for legacy workloads. However, many of the key assumptions that computer systems were originally based upon no longer hold, revealing inherent limitations in conventional abstractions that hinder the efficient use and adoption of promising new technologies. This is becoming increasingly concerning as efficiency and security demands on computer systems continue to grow, particularly in the era of ubiquitous AI and the end of Dennard Scaling.
In this talk, I will present Liquid Machine, a project spanning three generations of research and over fifteen years of efforts, aimed at addressing these foundational challenges. Traditional “push” approaches lead to fixed, often overprovisioned resources and inefficiencies in the computing stack that prevents us from fully exploiting modern and emerging technology. In contrast, Liquid Machine adopts a fundamentally new “pull” paradigm, enabling a dynamic, precise, and fully software-definable system stack adapting to actual application requirements and evolving technology capabilities. I will highlight two foundational works that make Liquid Machine possible. First, Liquid Silicon, a pull-based hardware platform that forges new ground across device physics/materials science, computer architecture, EDA/compiler, and chip implementation to achieve software-definable capabilities. Second, ENIAD, an innovative system software stack built on novel abstractions that complements Liquid Silicon and transforms the software-defined capability of the hardware into an end-to-end system gain. Through Liquid Machine, we illustrate how this paradigm shift from “push” to “pull” can illuminate new paths forward for our community and the future of computer engineering.
Bio: Dr. Jing (Jane) Li is the Eduardo D. Glandt Faculty Fellow and Associate Professor (with tenure) at the Department of Electrical and System Engineering (ESE) and the Department of Computer and Information Science (CIS) at the University of Pennsylvania. Dr. Li co-directs the CyberSavvy Research Center, a nationwide security initiative sponsored by DARPA. Previously she was the Dugald C. Jackson Assistant Professor at the University of Wisconsin–Madison and a faculty affiliate with the UW-Madison Computer Architecture group and Machine Learning group. She was one of the PIs in SRC JUMP center – Center for Research on Intelligent Storage and Processing-In-Memory (CRISP). She spent her early career at IBM T. J. Watson Research Center as a Research Staff Member after obtaining her PhD degree from Purdue University in 2009.
Host: Mingoo Seok