Over the decades, computers have changed dramatically, shrinking from room-sized machines to the tiny phones we carry in our pockets. But behind the scenes, computing principles have remained the same. That is something Akhilesh Jaiswal is trying to change.
“The basic computing paradigm essentially hasn’t changed since its inception. We are stuck with the same Boolean, or algebraic, computing people were performing ages ago. Now it’s just faster and better,” says Jaiswal, who joined the University of Wisconsin-Madison in summer 2023 as an assistant professor of electrical and computer engineering. “We’re asking whether we can develop new paradigms in computing because there are demands in terms the Internet-of-Things, AI and machine learning, real-time high-performance computing, and in other areas where the existing hardware computing paradigm conspicuously falls short.”
Jaiswal earned his bachelor’s degree from the Shri Guru Gobind Singhji Institute of Engineering and Technology in Nanded, India, before pursuing a master’s degree at the University of Minnesota. He received his PhD from Purdue University in 2019, then spent a year as a senior technology development engineer at the major semiconductor firm GlobalFoundries in Malta, New York. From there, he moved to the University of Southern California, where he held a joint position as a computer scientist with the Information Science Institute and as a research assistant professor in the school of engineering. There, he won several research awards, including the ISI Exploratory Research Award in 2020 and 2021 Keston Exploratory Research Award. He also has 29 patents and several patents pending to his credit.
One area Jaiswal is particularly interested in is extreme-edge computing. Currently devices such as cell phones more or less outsource much of their computing to the cloud. When you talk to a digital assistant like Siri, for instance, that information is processed in a data center hundreds or thousands of miles away. But with the next generation of smart devices, including smart clothing, sensors and medical equipment, connecting to the cloud won’t be fast or efficient enough—and using the traditional processing pathway, with separate memory and processing circuits, is too bulky and slow as well. Jaiswal proposes bringing the processing closer to the end user, or “edge,” through novel processing techniques.
One is called in-pixel processing. Digital cameras capture light, convert it into an electrical signal, and store it in memory. Any sort of processing, editing or application is offloaded to a processing chip or the cloud. But Jaiswal is working on a paradigm in which the camera is able to use its millions of pixels to actually do the computing and sensing.
“The question is, how do you bring computing closer to cameras and sensors?” he says. “The conventional answer is you add another chip to the board. But what we are doing is not adding another chip or sending data to the cloud. Instead, while the pixel is sensing, it’s also computing, making all these intelligent decisions on the fly.”
Jaiswal is also focusing on bringing processing power into computer memory. As AI advances, traditional computer architecture, in which the processor and memory are separated, will not be able to keep pace. Jaiswal is investigating computing-in-memory using static random-access memory or SRAM, a fast, efficient type of computer memory along with emerging non-volatile memory as well as optical memories that can read and write at ultra-fast speeds using laser sources.
He is also involved in neuromorphic computing, which mimics some of the efficient ways the brain works. For instance, the retina, or group of nerve cells at the back of the eye, can often perceive threats and cause the body to react before the brain realizes what is happening. Through his collaboration with retinal neuroscientists, Jaiswal is working to develop a camera with similar reactions. “So it is a brain-inspired solution to create what I’m calling the next generation of retina inspired cameras for autonomous systems and autonomous vehicles,” he says.
While he works at the edge of computing, Jaiswal says his time in the semiconductor foundry industry has given him a unique perspective, and he tries to pursue research with a practical side. “A lot of times when you look at great ideas, they rely on technologies that are not around to take them forward,” he says. “A lot of the projects I work on have commercial viability, as well as a straightforward roadmap to mass manufacturing today.”
Top photo by Joel Hallberg