Keeping in step with private industry, Los Alamos National Laboratory has seen a significant increase in resources dedicated to AI-specific projects, and an event hosted in the AI Technology Laboratory over the winter was a further development in advancing the Lab's AI for Mission — or ArtIMis — project. A team of about 30 members of the ArtIMis project participated in a hackathon under the direction of Earl Lawrence, the National Nuclear Security Administration’s models pillar lead for the Genesis Mission — a Department of Energy-wide AI effort — and a senior scientist in the Computing and Artificial Intelligence Division Office. The goal of the hackathon was to link the Universal Research and Scientific Agent (URSA), an agentic AI architecture built under ArtIMis, to other AI products produced through ArtIMis.
Agents and foundations
Agentic AI are systems that can be used to perform specified tasks without requiring much human involvement or input. URSA was developed to help scientists at the Lab answer questions about their research or complete duties that might otherwise be repetitive or time-consuming. Nathan DeBardeleben, a senior research scientist and co-principal investigator for ArtIMis, likens using URSA to a mentor giving a mentee a problem to solve. From the initial text query, he says a user might have to provide hints to guide URSA toward a desired outcome.
"It works toward a solution and then finally stops when it thinks it has met your criteria," DeBardeleben says. "A student is a good approximate — sometimes you'll think you defined the problem incorrectly to the student, other times the student stops before you wished they had, other times it finds something interesting and goes off on a bit of a tangent. You can pick up the pieces whenever you want with URSA at these checkpoints and steer it back on course."
URSA
URSA can summarize literature, write code, run scientific simulations, make plots, hypothesize new simulations or write papers as it works toward an answer. To produce these responses, URSA relies on a repository of accessible information and models, and to expand this library of knowledge, the hackathon participants integrated ArtIMis products known as "scientific foundation models" into URSA. These foundation models are pre-trained AI systems that have been unleashed on large Lab datasets and are now capable of identifying patterns to accomplish complex or new tasks for which they were not initially trained. By integrating these foundation models into URSA, the framework becomes more adept at answering a wider array of questions that Lab scientists might pose.
"As URSA becomes more capable, and as our agentic system can access the knowledge encapsulated in these big models, it becomes more able to solve the scientific problems we have, particularly for mission," Lawrence explains. "We're moving from the individual component stage to the ecosystem stage, and that makes everything a lot more useful to an actual scientist."
A common goal
With nearly 100 Lab employees contributing to the ArtIMis project, collaboration across the AI-focused capability-building teams (agentic, foundation models, and testing and evaluation) and scientific-application teams (multi-physics, fractures, chemical separation, and material discovery) is no easy task. Thus, Casleton identified a hackathon as an opportunity to get members of the different teams in one location at one time to build bonds across the different teams.
"A big thing that came out of it was just seeing who's doing the model building, who's doing the testing and evaluation," says Emily Casleton, a scientist and team leader of the ArtIMis testing and evaluation team. "Now, if anyone has a question about something, I know who to go to."
Casleton's team was charged with ensuring the foundational models were connected to URSA and running properly, and as the team lead, Casleton — as she put it — was the hackathon's "final boss."
"She said before everyone leaves today, she has to be able to run their model," Lawrence says. "I don't know if we quite achieved that, but regardless, everyone made such tremendous progress that I really think the productivity was quite high."
Contributing to Genesis
Both Lawrence and Casleton also noted that the hackathon can be instructive for the national goals outlined in the Genesis Mission. The process of making the foundational models accessible to URSA serves as a proxy for what is expected to happen on a national scale, and after the success of the hackathon, the transition from ArtIMis to Genesis should be easier.
"If we want someone at another lab or a large DOE-wide ecosystem to be able to access the things we developed, it's the exact same process," Lawrence says. "So, in my mind, that was always an ulterior motive. If we solve the problem and make sure our test and evaluation team can run stuff, we also contribute to the national initiative."
At a more insular level, the hackathon is a major step in providing more scientists with a robust AI tool for use across the Lab's mission spaces. In the future, Lawrence hopes to host a hackathon or a similar event every quarter, so the ArtIMis project can maintain a cohesive thrust and direction.
"There's a lot of hype and a lot of excitement around AI, particularly with Genesis, but it was nice to see the nuts and bolts of this all come together, and how scientists and AI experts can work side by side to solve real problems," Casleton says. "It's not just fluff, and that's really exciting for me."
LA-UR-26-22430
Contact
Media Relations | media_relations@lanl.gov
