Jason Simms, research computing manager at the College, is helping lead a $24.5 million National Science Foundation-funded initiative aimed at advancing AI-driven scientific research.

This past May, the National Science Foundation (NSF) awarded the Morgridge Institute for Research in Madison, Wis., a five-year grant to advance its project, Fabric for AI-Driven Science (FabAID), for which Simms serves as a co-principal investigator (co-PI). The award was part of a larger $83 million federal investment in projects seeking to reinforce American research infrastructure and improve opportunities for AI-driven research.

“I was very happy to be contacted by Brian Bockelman, who is the PI on the award from the Morgridge Institute,” Simms says.

It’s one of several NSF grants Simms has secured in recent years, notes Andrew Ruether, senior director, academic technology at Swarthmore, including one through the Office of Advanced Cyberinfrastructure program in 2024.

“Through his success in attracting research funding, his leadership in organizing our annual Research Computing at Smaller Institutions conference, and his strong support for faculty and students, Jason has established Swarthmore as a national leader in research computing at liberal arts colleges,” says Ruether.

Simms, who has experience serving small liberal arts institutions, says his role as co-PI of FabAID is to “engage with the smaller institutions nationally, an aspect of the project he finds particularly special. 

“Many of these grants don't take a position where smaller institutions are at the forefront,” he adds, “but this is different, and so it gives me a good opportunity to participate in a grant where that is the case.” 

Massive datasets are a problematic phenomenon looming over the current research landscape that FabAID aims to mitigate. 

“One of the core problems is that scientific challenges, scientific discoveries, and scientific research are increasingly requiring larger amounts of data,” says Simms. 

Sensitive instruments are capable of capturing more data with more precision, and the increasing complexity of research questions may demand bigger hauls caught with wider nets. This makes wrangling all that data more difficult, especially for smaller institutions with less access to higher computing power. Larger datasets may also be difficult to store locally, posing further challenges with data handling. 

Simms notes that both data wrangling and storage concerns trouble large institutions as well. 

“The challenge, even for very large institutions, is that sometimes, it is hard to store and work with those datasets locally,” he says. “ That's exacerbated by the fact that even at large institutions, there may not be access to sufficient amounts of compute resources to analyze those data locally.” 

Finally, a researcher on the West Coast interested in data stored in an unaffiliated institution on the East Coast, for instance, might not enjoy smooth and easy access. Data distributed across various databases could create administrative hurdles. 

The FabAID project aims to reckon with these concerns by weaving together an organized national fabric that connects data and computing power, ensuring any institution of any size across the country can smoothly tap into it for academic use. It also seeks to provide AI tools to ameliorate the research process. 

FabAID’s streamlined nature may greatly improve administrative accessibility. 

“One of the things that we want to do with FabAID is to create this fabric that seamlessly connects the local with the national, or with the regional, or with different scales of analyses,” says Simms. “Researchers and professors can use a common set of tools whether they are working locally or whether they are using huge amounts of distributed resources all throughout the country.” 

In fact, AI tools play a critical role.

“At a smaller institution, especially, the researcher is often teaching a full load of classes and doesn’t necessarily have access to a postdoc [postdoctoral researcher],” says Simms, “and they’re often reliant on undergraduate students to kind of be their research assistants.”

FabAID can provide such professors with AI tools to assist them with their academic work. Additionally, by caching data, FabAID aims to ensure datasets are able to be stored physically closer to more researchers. 

“We’re improving the tools to put these data on certain kinds of servers, either locally, regionally, or at a larger institution so copies of those data get replicated all throughout the country,” Simms says, noting that wrangling data is more convenient if it can be done closer to home.

Research, Simms notes, has historically been pursued by faculty at larger institutions with greater access to information and resources. He hopes FabAID could alter that narrative. 

“What I would hope is that, through projects like FabAID, we break down some of those barriers, and we make scientific research and teaching using these resources available to anyone, regardless of where you are,” says Simms. “I would like to hope that we contribute to making it possible to contribute to science and to teaching.” 

“We don't want the type of institution you're at and the local tools that you have to be a barrier to contributing to scientific research or to teaching your students in very applicable ways.”