Argonne: Where AI research meets education and training

Last September, in the Chicago suburb of Lemont, Ill., Argonne National Laboratory hosted its first AI STEM Education Summit. More than 180 educators from high schools, community colleges, and universities; STEM administrators; and experts in various disciplines convened at "One Ecosystem, Many Pathways-Building an AI-Ready STEM Workforce" to discuss how artificial intelligence is reshaping STEM-related industries, including the implications for the nuclear engineering classroom and workforce.

Among those who spoke at the summit were leaders from Argonne, computer technology company Intel, biotechnology firm Genentech, Chicago Public Schools, the Illinois Mathematics and Science Academy, Illinois Institute of Technology, Waukesha County Technical College, Chicago State University, and the nonprofit organization Education Development Center.

Like a number of other Department of Energy facilities, Argonne is at the forefront of some of today's most sophisticated AI technologies, including in such areas as modernizing energy infrastructure, extending battery life for cars, designing new materials, researching medical advances, creating more predictive climate models, and developing automated experiments. A key part of much of this research is Argonne's Aurora, one of the world's first and fastest exascale supercomputers, capable of performing at least one exaflop, defined as a quintillion (10¹⁸) calculations per second.

Argonne is also home to the Mechanisms Engineering Test Loop (METL), where research incorporates AI into immersive technologies such as augmented reality (AR, the overlay of digital information onto real-world environments in real time) and extended reality (XR, a broad term for immersive technologies that create interactive experiences). METL is the nation's largest liquid metal test facility dedicated to developing small- to intermediate-scale components for advanced sodium-cooled fast nuclear reactors. The METL team conceptualizes, fabricates, and demonstrates equipment and instrumentation for such reactors.

According to Derek Kultgen, who at the time of the September summit was the group leader and principal mechanical engineer for METL, "The main point regarding XR for nuclear workforce training is that, using the METL facility, we were able to demonstrate XR applications throughout the life cycle of a nuclear facility. Specific examples included the acceleration of fabrication and minimization of errors during construction, communication of new designs, bringing a control room to any room using a Meta Quest headset, and digitizing operations."

Summit highlights

The AI STEM Education Summit featured speeches, panel discussions, and interactive breakout sessions in which participants exchanged ideas on integrating AI into STEM classrooms and the workforce. Leaders from research, education, and business shared real-world examples of the impact of AI on their fields.

In his remarks to summit participants, Argonne Director Paul Kearns said, "Preparing the next generation of scientists, engineers, and innovators is one of Argonne's highest priorities. Our ability to keep changing the world with science depends on partnerships like those we are building here today."

Among the business leaders who addressed the summit was Byron Reese, a Texas-based tech entrepreneur. Reese, who holds several technology patents, delivered a keynote speech that described how AI is accelerating scientific and technology discoveries.

Rajeev Thakur, deputy director of Argonne's Data Science and Learning Division, also discussed the acceleration of discoveries through the use of AI, including in the areas of energy, national security, and human health. "AI will help us solve major challenges, but its real legacy will come from the discoveries and innovations we cannot yet imagine-made possible by the people who know how to harness it. The future of STEM will belong to those who are ready, and it is our responsibility to ensure students are prepared to lead the way," he said.

Meridith Bruozas, the director of institutional partnerships at Argonne, noted that "Argonne researchers are demonstrating how AI is transforming science itself. This summit is a chance not only to share our work, but also to learn from others who are advancing AI in education and research. By bringing those insights to educators, we can help classrooms mirror real-world innovation and give students the skills and perspective they need to shape the future."

Genesis Mission and AI Jam

Bruozas told Nuclear News, "Artificial intelligence is revolutionizing and accelerating scientific discovery and transforming how science is conducted. To remain competitive, we must develop an AI-ready STEM workforce. To support this and in alignment with DOE's Genesis Mission, Argonne is convening education leaders through events and symposiums like the AI STEM Education Summit and the AI STEM Education Jam."

The Genesis Mission was established by presidential executive order last November with the goal of developing an "integrated platform that connects the world's supercomputers, experimental facilities, AI systems, and unique datasets across every major scientific domain to double the productivity and impact of American research and innovation within a decade." The AI STEM Education Jam (also referred to as the 1,000 Scientist AI Jam), which was held on February 28, was a hybrid in-person/online event held at Argonne, Brookhaven, Idaho, Lawrence Berkeley, Lawrence Livermore, Los Alamos, Oak Ridge, Pacific Northwest, and Princeton Plasma Physics National Laboratories. Participants explored advanced AI models at this first of several planned "jam sessions."

Bruozas said, "These events raise awareness, provide hands-on learning experiences, and build a network of education leaders committed to integrating AI into STEM classrooms and future learning experiences. These events also highlight the need for STEM education to evolve and are essential to preparing the next generation of AI-ready STEM talent."

The September summit featured a lot of information about Argonne's use of AI, XR, machine learning, data science, and high-performance computing to drive scientific breakthroughs.

Aurora, which was built in partnership with Intel and Hewlett Packard Enterprise, was made available for use by researchers around the world in January 2025 and is powering many of those breakthroughs. Located at the Argonne Leadership Computing Facility, it is equipped with 63,744 graphics processing units and 84,992 network endpoints. It spans eight rows of refrigerator-sized cabinets with 166 computer racks, weighs 600 tons, covers 10,000 square feet, and contains 300 miles of networking cables. Aurora consumes as much energy as thousands of homes and requires a cooling system with 44,000 gallons of chilled water.

METL research

While Aurora must be kept cool, Argonne's METL facility, established in 2018, holds 750 gallons of reactor-grade sodium that can reach 650°C. It has more than 1,000 sensors that collect diagnostic data on the testing of components and systems for liquid metal reactor technologies, including fuel-handling systems, self-actuated control and shutdown systems, sensors and instrumentation, and in-service inspection and repair technologies.

In addition to its wide-ranging R&D applications, METL can be used to collect operational data in real time and advance the training and education of the nuclear workforce.

A 2023 Argonne report, Deploying Extended Reality (XR) for Digital Operations and Maintenance at the Mechanisms Engineering Test Loop (METL), looked at how AI and XR technologies can be used in the training and work activities of the next generation of nuclear professionals. Kultgen and his fellow researchers found, for example, that the digital scanning of an engineering drawing of a cask used for inserting and retrieving equipment in METL procedures "allowed the user to augment the equipment at full-scale during a design review, enabling engineers, designers, and technicians to 'see' and 'walk-around/through' the device to help identify interferences, operational challenges, etc., prior to fabrication."

Kultgen described another example "using a Meta Quest [virtual reality] headset" where "operators were immersed in a 'metaverse,' which allowed them to walk around a virtual room. While in the virtual room, they could populate as many virtual monitors as needed to display real-time METL data. This would allow operators to monitor large facilities from anywhere and also negate the need for the tens to hundreds of physical monitors commonly seen in reactor control rooms."

Kultgen and his team created a digital version of METL, about which he said, "Essentially any variable can be populated into it to provide a color gradient of the status or performance of the entire plant, allowing an operator to have a high-level view of the facility and identify potential problem areas. They can also walk around the virtual plant for closer inspection. This is expected to be very valuable in a microreactor or small modular reactor, where operators are overseeing many plants."

What the future could hold

Argonne's AI STEM Education Summit and Education Jam, together with the work being done with the Aurora exascale supercomputer and the METL facility, reveal the many ways in which AI-related technologies are having an impact on nuclear education and research and on the nuclear workforce. So what might the future hold?

Kultgen shared some of his thoughts on this matter: "If someone doesn't understand a complex topic, we often don't question whether the information was presented in a digestible way. AI could be very useful from an education or training aspect, in the fact that it could provide many explanations from different viewpoints or provide nearly endless examples. Topics like nuclear energy are difficult to learn, because most people don't have a power plant in their 'backyard.' Using XR and other immersive environments can bring the plant to the classroom to help students fully understand the scale and complexity required in its operation."

Nuclear workforce training is likely to benefit substantially from the continued development of AI technologies. "AI is already quite proficient at directing users to additional resources, such as providing names of good books on particular subjects," Kultgen said. "In the near future, I would expect AI to provide rapid access to data as well, including responses to such directions as 'Show me a graph of fuel temperature vs. time after removing a control rod in (insert control rod location) for reactor (insert reactor name) given a cold-start condition.'"