Bringing AI to Scientists

KJ Schmidt (MSAI ‘20) sees so much potential for artificial intelligence (AI) in the sciences.  

Schmidt is an AI research scientist and product manager at Globus Labs at the University of Chicago. She also has a joint appointment at Argonne National Lab, the nation’s largest federally funded R&D center.   
In her dual roles, she leans on her experience in Northwestern Engineering's Master of Science in Artificial Intelligence (MSAI) program to help scientists do more with AI.

“The most exciting part about the work I do is the impact we have on science and the emphasis on creating tools that make science more open,” she said. “There's always more to learn. There's new research and new applications all the time, but there are so many questions that remain or pop up as you go.”

Many of those questions relate to the large language models at the center of many AI applications. For all of its potential, Schmidt still likes to remind people that AI isn't leading a new revolution — it’s being guided by humans, with all their fallibility and biases.

“What worries me about current conversations about AI is that people put a lot of faith into the model outputs,” she said. “The conversations saying that models ‘know’ things? They don't. Under the hood it's all math.”

Math wasn’t Schmidt’s strongest skill when she joined the MSAI program. She saw herself as an accomplished coder, but was challenged by the higher mathematical principles underpinning AI development.

Growing in that area was a turning point for Schmidt as she realized the value of collaboration.

“Everyone has different strengths,” she said. “I found that some people had a math background but had just learned how to code. When I felt out of my depth looking at these complex equations, others felt out of their depth when building software applications.”

It’s a lesson that continues to be helpful in her work at both Globus Labs and Argonne National Lab. Her role focuses on developing software tools for scientists that help them more easily apply AI to their research. She also uses communication skills to write website copy, articles, and slide presentations to spread the word about what the two labs are discovering.

One of Schmidt's main projects is Foundry-ML, an open-source machine learning platform for scientists. Researchers at Argonne and the University of Toronto made the largest available set of quantum Monte Carlo calculations, and that dataset is now on Foundry-ML.

"That makes it significantly easier for other researchers to access, understand, and use," Schmidt said. "Having this data available helps researchers explore the chemical space with far higher accuracy."

In all her efforts, Schmidt’s focus is using AI to promote social and scientific good. It’s what got her involved in AI in the first place.

“When I was a software engineer, there was a company I heard about that was making prosthetics using AI so the user would be able to touch and move and feel with the prosthetic limb,” she said. “This blew my mind. I was so inspired to learn more about AI and figure out how I could be a part of this technology and use it for good.”  

The MSAI program gave her the foundation she needed to help steer AI in that direction. She credits a range of classes as being vital to her growth, including Computation, Ethics, and Society; Natural Language Processing; Statistical Language Modeling; and AI in Law.

She recently returned to campus to talk with current MSAI students, sharing the most important lessons she learned in the program.

"You can do hard things," Schmidt said. "AI can be daunting. Find a way to break things down into pieces you understand, and build up from there."