W&M awarded $1M DOE grant to advance AI-assisted detector design, optimization

William & Mary has received a new $1 million grant from the U.S. Department of Energy to lead the next phase of AID2E.

Engineering model of the future ePIC detector at the Electon-Ion Collider currently being built at Brookhaven National Lab (Image courtesy of Sean Preins/VIRTUE)

The following story originally appeared on the website for W&M's School of Computing, Data Sciences & Physics. - Ed.

William & Mary has received a new $1 million grant from the U.S. Department of Energy to lead the next phase of AID2E, an ambitious effort to integrate artificial intelligence into the design and optimization of large-scale physics experiments with multiple detector systems.

AID2E - an AI-assisted Detector Design and Optimization framework for Experiments at the Electron-Ion Collider and beyond - addresses a key challenge shared across modern nuclear and particle physics experiments, designing and optimizing complex experimental systems with many interconnected parameters and engineering considerations to support multiple physics goals.

The project develops an AI-assisted framework to support the design and optimization of detectors for future experiments at the Electron-Ion Collider (EIC), a flagship U.S. Nuclear Physics facility being built at Brookhaven National Laboratory and developed in partnership with Thomas Jefferson National Accelerator Facility (Jefferson Lab) and a broad scientific collaboration.

Once completed, the EIC will enable detailed studies of the internal structure of nucleons and nuclei, advancing our understanding of how quarks and gluons give rise to the mass and structure of visible matter, with broader scientific and technological impacts.

The initial phase of the project started in 2023, with AID2E demonstrating the feasibility of using AI to inform and support detector design studies. The funding for this next phase is part of a broader $1.45 million award that includes Brookhaven National Lab, Jefferson Lab, the Catholic University of America and Duke University.

"AI-assisted detector design and optimization will allow the EIC and other large-scale experiments to achieve higher performance, lower operational costs, faster design cycles, and more ambitious physics goals than were previously possible," said Cristiano Fanelli, a W&M associate professor of data science and the principal investigator for the project.

AID2E integrates advanced optimization techniques with modern, data-driven workload management systems to support complex detector design and optimization studies at scale. The renewed effort will introduce large language model agents to help streamline and coordinate these workflows, while ensuring that researcher oversight remains central.

Solving a critical challenge in modern physics experiments

Imagine trying to design the world's most complicated scientific camera - an instrument designed to precisely measure particles and radiation - one that has hundreds of interdependent parts. Changing one thing, like the shape of a sensor or the material used in a layer, can change how the whole system works. Historically, these highly advanced systems were optimized largely by hand, making full exploration of the design space virtually impossible.

AID2E uses artificial intelligence to help with this enormous puzzle.

By exploring design spaces far larger and more complex than humans can navigate, AID2E can reveal non-intuitive configurations, can deliver simultaneous performance gains, sharpen particle identification and tracking precision, extend usable acceptance, and reduce material budget or cost while still meeting or exceeding the experiment's physics goals.

"Detector complexity, the computational cost of large-scale simulations, and growing experimental demands have surpassed what manual approaches can handle," stated Fanelli. "Modern AI methods, supported by the vastly expanded computing capabilities of recent years, have matured into reliable and scalable solutions."

New possibilities for detectors and beyond

AID2E supports applications across experiments and scientific domains.

It has already been successfully applied to alignment studies for the CLAS12 RICH detector at Jefferson Lab and is now being applied to materials R&D, including the design of a new fiber-reinforced aerogel. This extremely light, porous material is being engineered to achieve greater strength and improved optical quality, both of which are critical for advanced detectors and related technologies.

"AID2E is agnostic to what is being optimized, making its applications far broader than the EIC," stated Fanelli. "It can support the design of future detectors, improve alignment and calibration in ongoing experiments, and tackle a wide range of optimization tasks."

Over the next two years, the AID2E team aims to integrate the framework into EIC simulation campaigns, expand its use across additional subdetector systems, and extend its demonstrated successes in alignment, materials research and workflow automation across laboratories nationwide.

National leadership in AI-for-science

AID2E positions W&M at the center of a nationally visible, multi-institution effort that bridges physics, AI, data science and advanced computing.

AID2E collaborates with the EIC's ePIC Collaboration, which includes more than 180 institutions across 25 countries. Leading this DOE-funded AID2E project showcases W&M's strength in developing advanced AI methods, integrating AI with large-scale scientific workflows and delivering tools across major national and international research facilities.

The project provides graduate and undergraduate students the opportunity to participate directly in AI-for-science research, contribute to software used at national labs, and engage with the EIC community, gaining experience in modern AI methods, scientific computing, workflow orchestration, and real detector applications.

"For many students it will be their first involvement in a national-scale scientific project, preparing them for competitive careers in AI, data science, and experimental physics," added Fanelli.

The project also positions W&M to play a larger role at major DOE facilities, providing AI and software capabilities that are increasingly essential across all phases of an experiment.

"This project reflects the strength of university-laboratory partnerships in advancing AI-enabled science," said Fanelli. "AID2E brings together complementary expertise in physics, detectors, AI, software, and computing across institutions, creating a collaborative environment where ideas can be developed, tested, and scaled using the robust computing and workflow infrastructure available through national laboratories."

Over the next decade, Fanelli believes scientific workflows will become increasingly AI-assisted, with systems capable of interacting with scientists through natural language and coordinating complex, large-scale computing. This aligns with the national vision of AI-enabled science to accelerate scientific advancement.

"AID2E shows how AI, integrated into scientific workflows, can assist scientists across critical stages of an experiment - from design through optimization during commissioning and operations," said Fanelli. "These AI capabilities are becoming essential as experiments grow in scale and complexity, helping to streamline and modernize how large-scale experiments are designed and operated."

Editor's Note: William & Mary is committed to preparing students for data-rich environments and an AI-driven world through thoughtful leadership and human-centered innovation. This vision is taking shape in the new School of Computing, Data Sciences & Physics (CDSP) in collaboration with the entire campus. CDSP integrates AI tools into daily work, including news writing. The CDSP communications team used OpenAI's ChatGPT to assist in building this article. The team then reviewed and edited the article before publication

Randy Ready, School of Computing, Data Sciences & Physics

• share by email
• share on Facebook
• share on X
• share on LinkedIn

• share by email
• share on Facebook
• share on X
• share on LinkedIn

Related Stories