The U.S. Department of Energy has selected North Carolina State University to lead three projects designed to use artificial intelligence to accelerate discovery across high-performance computing, cybersecurity and astrophysics. The awards are part of the DOE’s Genesis Mission, a national initiative intended to increase scientific productivity by combining advanced computing, automated reasoning and large-scale research data. NC State is also participating in two additional projects focused on nuclear energy, including a major effort supported by $60 million over three years.
The Genesis Mission is built around the idea that artificial intelligence can become a powerful scientific instrument rather than merely a tool for analyzing completed experiments. AI systems can help researchers write and optimize computer code, identify unusual patterns in complex data, monitor scientific workflows and approximate physical processes that are too expensive to calculate directly. The first round of awards includes short-term Phase I projects, each funded at less than $1 million and intended to be completed in under a year. Projects that demonstrate strong results may advance to larger Phase II awards in future funding cycles.
“NC State has a deep bench of research experience in AI development, application and integration,” says Krista Walton, the university’s vice chancellor for research and innovation. She says the university’s leadership in three projects will help develop technologies needed for the “post-AI era,” while ensuring that artificial intelligence is used safely in scientific research. The projects span both fundamental science and the underlying computing infrastructure required to make advanced AI systems practical for researchers who may not be specialists in machine learning.
One NC State-led project, headed by Michela Becchi, assistant professor of electrical and computer engineering, will focus on making emerging computing hardware easier for scientists to use. Modern AI accelerators, including specialized processors originally designed for neural-network calculations, can perform certain operations far more efficiently than conventional central processing units. However, programming these devices often requires highly specialized knowledge, and scientific software may need substantial redesign before it can run effectively on them.
Becchi’s team plans to develop an AI-powered programming framework capable of generating, debugging and optimizing code for these next-generation accelerators. Such a system could analyze scientific programs, identify sections suitable for parallel execution, translate them into hardware-specific instructions and detect errors that would otherwise require extensive manual testing. It may also help researchers tune code to balance speed, memory use and numerical accuracy. The project is being conducted with Sheng Di and Bogdan Nicolae of Argonne National Laboratory, where advanced computing systems are used for research ranging from materials science to climate modeling.
A second project, led by Xiaorui Liu, assistant professor of computer science at NC State, will address the security of agentic AI systems. Unlike conventional software that performs narrowly defined tasks, agentic AI can plan and execute multistep workflows, call external tools, retrieve information and make decisions based on intermediate results. In scientific settings, these systems could automate simulations, data analysis and literature searches, but their autonomy also creates new opportunities for malicious interference.
The project will investigate methods for detecting and preventing adversarial attacks against automated scientific workflows. An attacker might manipulate input data, exploit weaknesses in an AI model or alter an intermediate computation so that the final result appears plausible while being scientifically false. Other attacks could cause an AI agent to reveal confidential information or follow unauthorized instructions. The researchers aim to develop protections that preserve the usefulness of agentic systems while enabling scientists to audit decisions, verify outputs and identify compromised results. Collaborators include researchers from the University of Tennessee, Knoxville, Michigan State University and Depict Bio.
The third NC State-led project will apply AI to some of the most energetic and difficult-to-model events in the universe. Gail McLaughlin, Distinguished University Professor of Physics, is leading an effort to embed AI techniques within large-scale simulations of neutron star mergers and supernovae. These events involve extreme densities, temperatures, magnetic fields and nuclear reactions, producing physical interactions that can be prohibitively expensive to calculate in full detail.
AI models could help approximate portions of those calculations while preserving the most important physical relationships. In practice, researchers may use machine-learning surrogates to represent processes that require enormous computational resources, allowing simulations to explore more conditions or run at higher resolution. The project will investigate how such methods can improve understanding of the cosmic origin of the elements and the role of neutrinos in stellar explosions. Neutrinos are nearly massless particles that interact weakly with matter, yet they can transport vast amounts of energy during a supernova and influence the formation of heavy elements. The collaboration includes scientists from the University of Tennessee, Pennsylvania State University, the University of Notre Dame, the University of California, Berkeley, and Oak Ridge National Laboratory.
NC State is also a partner on two Genesis Mission projects involving nuclear energy. One Phase I project, led by Texas A&M University, will explore the use of AI to assess and document reactor safety in ways that support human review. Nuclear safety analysis requires the examination of large volumes of technical information, operating conditions and potential failure scenarios. AI could help organize evidence, identify inconsistencies and produce documentation, but final judgments would remain subject to expert oversight. Xu Wu, associate professor of nuclear engineering at NC State, is the university’s lead on the project.
The second nuclear initiative, known as the Prometheus project, is the only Phase II award announced in the initial Genesis Mission round. Led by Idaho National Laboratory, it will receive $60 million over three years to help accelerate the deployment of safe and affordable nuclear power infrastructure. The project brings together expertise intended to address technical and engineering barriers that can slow the construction and operation of advanced nuclear systems. NC State participants include Wu, Abhinav Gupta, professor of civil, construction and environmental engineering, and Kevin Han, associate professor in the same department and Edward I. Weisiger Distinguished Scholar. Together, the projects position NC State at the intersection of artificial intelligence, scientific computing, cybersecurity, astrophysics and nuclear engineering—fields where faster computation and more reliable automation could reshape how research is performed.
Subject of Research: Artificial intelligence applications in high-performance computing, cybersecurity, astrophysics and nuclear energy
Web References: U.S. Department of Energy Genesis Mission announcement; Idaho National Laboratory Prometheus project announcement
Keywords
Artificial intelligence, Genesis Mission, North Carolina State University, high-performance computing, AI accelerators, cybersecurity, agentic AI, adversarial attacks, astrophysics, neutron star mergers, supernovae, neutrinos, nuclear energy, scientific computing, DOE, Idaho National Laboratory
Tags: advanced computing and automated reasoningAI as a scientific instrumentAI for cybersecurity and astrophysicsAI optimization of computer codeAI-driven research data analysisArtificial intelligence in scientific discoveryDOE funding for AI and scientific innovationDOE Genesis Mission projectshigh-performance computing for researchlarge-scale research data utilizationNC State University research leadershipnuclear energy research funding
