Prof. Boris Murmann
Professor of Electrical Engineering, Stanford University
Boris Murmann is a Professor of Electrical Engineering at Stanford University. He joined Stanford in 2004 after completing his Ph.D. degree in electrical engineering at the University of California, Berkeley in 2003. From 1994 to 1997, he was with Neutron Microelectronics, Germany, where he developed low-power and smart-power ASICs in automotive CMOS technology. Since 2004, he has worked as a consultant with numerous Silicon Valley companies. Dr. Murmann’s research interests are in mixed-signal integrated circuit design, with special emphasis on sensor interfaces, data converters and embedded machine learning. In 2008, he was a co-recipient of the Best Student Paper Award at the VLSI Circuits Symposium and a recipient of the Best Invited Paper Award at the IEEE Custom Integrated Circuits Conference (CICC). He received the Agilent Early Career Professor Award in 2009 and the Friedrich Wilhelm Bessel Research Award in 2012. He has served as an Associate Editor of the IEEE Journal of Solid-State Circuits, an AdCom member and Distinguished Lecturer of the IEEE Solid-State Circuits Society (SSCS), as well as the Data Converter Subcommittee Chair and the Technical Program Chair of the IEEE International Solid-State Circuits Conference (ISSCC) and the inaugural tinyML Research Symposium. He currently chairs the IEEE SSCS future directions committee (SSCD). He is a Fellow of the IEEE.
Abstract
Over the past decade, machine learning algorithms have been deployed in many cloud-centric applications. However, as the application space continues to grow, various tinyML algorithms are now embedded “closer to the sensor” and in wearable devices, eliminating the latency, privacy and energy penalties associated with cloud access. While the first wave of tinyML hardware will be largely based on conventional, accelerator-enhanced compute platforms, it is foreseeable that the quest for ultra-low power will drive us toward higher levels of customization. Within this context, I will review challenges and opportunities that arise in the design of full custom tinyML systems. The discussion will examine asymptotic limits, as well as opportunities in circuit design, highlighting analog feature extraction and mixed-signal computing. In addition, we will discuss challenges that arise due to the complex design trade-off space spanned by software, algorithm, and hardware metrics.