Chips for AI and AI for Chips: The Symbiotic Future of Computer Hardware

Abstract: Artificial Intelligence (AI) has revolutionized industries across the board, and chip design is no exception. As AI continues to grow in importance, the demand for more efficient computing systems has only intensified. This has led to the increase in chips for AI – computer chips and systems specifically designed for high efficiency on AI workloads. At the same time, more powerful AI models and greater processing efficiency have led to a renaissance in electronic design automation. Machine learning algorithms are paving the way towards smarter, faster chip design. In this talk, I will discuss my recent work towards both of these fronts. I will cover enabling Large Language Models (LLMs) for physical chip design tasks, using Reinforcement Learning (RL) to enable faster convergence in chip routing, and designing co-packaged optical systems for faster AI training.

Bio: Austin Rovinski is an assistant professor in the Electrical and Computer Engineering department at New York University’s Tandon School of Engineering. He received his Ph.D., M.S.E., and B.S.E. degrees all from the University of Michigan – Ann Arbor. Prior to joining NYU in Fall 2023, Austin spent a year as a postdoc at Cornell University. Austin’s work focuses on the intersection of computer architecture, VLSI, and EDA, having published at top venues such as ASPLOS, VLSI, and ICCAD. Austin has received an IEEE Micro Top Picks honor (2015), Dwight F. Benton Fellowship (2016), NSF Graduate Research Fellowship honorable mentions (2017, 2018), and best paper nomination at LAD (2024). Austin is also a founding member of The OpenROAD Project and an active contributor.