UMD to Become Hub for Industrial AI
Jay Lee, a major force in pioneering Industrial AI, joined the University of Maryland’s (UMD) mechanical engineering faculty this fall as Clark Distinguished Chair. Among his key aims at UMD: creating opportunities for both undergraduate and graduate studies to gain experience in the field.
As its name suggests, Industrial AI is concerned primarily with applications within industry. Lee defines it as a “systematic discipline which focuses on developing, validating, and deploying various machine learning algorithms for industrial applications with sustainable performance.” Lee is launching an Industrial AI Center at the university, with operations expected to begin in September 2023. In conjunction with the center, Lee plans to establish an Industrial AI Foundry, featuring more than 100 industrial data sets, in order to scale up AI education and research.
Lee is also introducing Industrial AI to undergraduate and graduate students through courses he teaches in the mechanical engineering department, thus updating the existing curriculum to take into account the growing interest in machine learning, AI, and data.
“AI should not belong just to advanced researchers,” Lee said. “I want to bring it to the undergraduate level, starting in the sophomore year. Traditionally, we use a physics-based approach to engineering education, with every student taking physics, chemistry, and calculus. But our society today needs both a physics-based and a data-centric approach.”
Students enrolled in Lee’s Introduction to Data-Centric Engineering course will learn AI, machine learning, and data science fundamentals, and gain experience using tools such as MATLAB. The course is designed to complement existing programs within the department, Lee said.
“Historically, AI has focused on data in and of itself—that is, on bringing in large amounts of data and then using it to train the machine. In the context of industry, though, we need to do more than that. We need to ensure the data is relevant to the context, and that it’s data of sufficiently high quality."
While AI education comes in many flavors, the Industrial AI approach advocated by Lee has three major elements, termed Data, Domain, and Discipline. One of Lee’s key insights is that data gathering should be guided by the needs of a particular domain—for example, automotive manufacturing or electrical engineering. Discipline–that is, quality control of the data–must also be incorporated into the framework to ensure good results. Machines that train themselves on spurious or poisoned data are unlikely to perform well.
“Historically, AI has focused on data in and of itself—that is, on bringing in large amounts of data and then using it to train the machine,” Lee said. “In the context of industry, though, we need to do more than that. We need to ensure the data is relevant to the context, and that it’s data of sufficiently high quality. This is what I mean by the ‘three Ds,’ and it’s what I teach my students.”
In addition to teaching courses, Lee will also be providing research opportunities for students, including undergraduates, in the Industrial AI Center–opportunities that will not only help them build expertise, but also connect them with major companies with an interest in Industrial AI.
Prior to joining the mechanical engineering department at UMD, Lee was a faculty member for 18 years at the University of Cincinnati, where he was an Ohio Eminent Scholar and L.W. Scott Alter Chair, as well as a University Distinguished Professor. He initially founded an Industrial AI Center at Cincinnati.
With Lee on the faculty, UMD’s Clark School of Engineering now becomes one of the few engineering colleges in the nation where students can study and gain hands-on experience with Industrial AI, working under a mentor who pioneered the field. Lee’s 2020 book Industrial AI: Applications with Sustainable Performance is considered a seminal work on the topic.
In addition to his academic work, Lee is a member of the World Economic Forum’s Global Future Council on Advanced Manufacturing and Value Chains. He also cofounded the company Predictronics, with customers that include companies such as Canon, Coca-Cola, Epson, Hibachi, Nissan, and Toyota.
Published February 28, 2023