A Hopkins-led research team has been awarded $750,000 in funding through Phase I of the U.S. Department of Energy’s (DOE) Genesis Mission, a national initiative bringing together laboratories, universities, and industry partners to develop AI-enabled scientific workflows to accelerate scientific breakthroughs in energy, discovery science, and national security.
According to the DOE, the 278 teams selected for this phase will develop new AI models and frameworks to address some of the nation’s most pressing challenges in energy, scientific, and engineering challenges. The initial awards provide support for nine months, with awardees eligible to compete for Phase II funding of $6–15 million over an additional three years.
Ziyang Li, an assistant professor of computer science and a member of Johns Hopkins’ Information Security (ISI) and Data Science and AI Institutes (DSAI), is the principal investigator of the awarded project “Hierarchical Neuro-Symbolic Synthesis of Verifiable Computational Physics Code for Scientific Discovery.” Fellow CS faculty member Yinzhi Cao, the technical director of the ISI and a member of both the DSAI and the university’s Institute for Assured Autonomy, joins him as a co-investigator, with additional support provided by Nengkun Yu of Stony Brook University and Meifeng Lin of Brookhaven National Laboratory.
The goal of their project is to transform high-performance computing software development from an expert craft requiring years of specialized knowledge into a structured, reproducible, AI-assisted scientific workflow.
“Scientific breakthroughs increasingly depend on sophisticated simulation software, but developing that software often takes years,” says Li. “Our goal is to build AI systems that not only generate scientific code, but also verify that it faithfully reflects the underlying physics. By combining AI with formal reasoning, we hope to make scientific computing faster, more reliable, and more accessible to researchers.”
Initially, the project will curate high-quality data for verified computational physics code and focus on applications in high-performance computing for particle physics before extending to other areas of computational science. Ultimately, this work could enable scientists to move more quickly from new scientific ideas to reliable computational tools, thus accelerating discoveries in energy, materials science, and and physics.