Every Thursday, Eric and Wendy Schmidt AI in Science Fellows gather—neuroscientists and pathologists breaking bread with climate scientists, ecologists, astrophysicists and particle physicists—to share research problems, learn from faculty experts, and collaborate across disciplinary boundaries. This weekly event, the “Unseminar Series,” is an important forum for Fellows and a hub where they connect directly with faculty, problem-solve together, and shape their own professional development. 

Launched in January 2023, the Schmidt AI in Science Fellowship brings together postdoctoral researchers from diverse fields who are applying, or interested in applying, AI and machine learning (ML) techniques to advance scientific research. The Unseminar combines some of the program’s greatest strengths: a community of pioneering interdisciplinary researchers, opportunities for professional development, and robust mentorship.

Collaborating Across Scientific Domains

Schmidt Fellow and neuroscientist Ramanujan Srinath was searching for an AI model that captures why humans can recognize that we’ve seen something before, but struggle to remember where we’ve seen it.

Srinath presented his conundrum at Unseminar, and his fellow researchers posed questions and possibilities. “Another Fellow mentioned Hopfield networks,” Srinath recalls, “and it sparked an idea.”  When he attended a subsequent Unseminar session on reinforcement learning, he realized he could try that approach as well—and it ultimately proved a better fit for behavioral data from neuroscientific experiments. Over the next year and a half, insights from these discussions evolved into a new line of experimentation for Srinath.

These informal open problems sessions, where Fellows bring challenges to brainstorm solutions as a group, are a core component of the Unseminar. As these researchers integrate AI and ML into their respective domains to forge new ground, the opportunity for interdisciplinary cross-pollination allows Fellows to benefit from methods developed in one domain that translate, sometimes unexpectedly, to others.

“One of the most valuable aspects of being in this room is the interdisciplinary perspective,” said Rui Ding, a third-year Fellow. “Someone working in a completely different field might frame a problem in a way that suddenly makes your own challenge click. That’s incredibly valuable when you’re working at these frontiers applying AI to scientific research.”

“For instance,” Ding explained, “my research focuses on functional devices in chemistry and materials science. During an Unseminar session, another fellow studying neuroscience presented the idea of using reinforcement learning to efficiently deploy AI agents. Our discussion inspired me to leverage that methodology in my own work, and led to improved results I wouldn’t have anticipated.”

Beyond solving immediate research problems, these Thursday gatherings build community among early-career researchers pioneering the application of AI across scientific domains. Fellows learn not just from faculty experts but from each other, sharing approaches, troubleshooting techniques, and perspectives that emerge from their diverse training and research questions.

Learning from Expert Faculty

AI tutorials like the one that gave Srinath the reinforcement learning insight he needed also demonstrate the engagement of faculty leadership.

Last October, when Fellows expressed challenges articulating their research questions across disciplines to their AI Skills Mentors, the DSI’s Faculty Director of AI Rebecca Willett led a hands-on communications workshop. Willett had Fellows practice explaining their research to one another and provided real-time feedback tailored to each pair’s specific strengths and challenges.

“This is a training program, and I really believe in developing these scholars to be future leaders,” Willett said. “The workshop gave them an opportunity to practice articulating their research across disciplinary boundaries in an encouraging environment, while building community and supporting one another’s development.”

Other faculty have brought similarly engaged approaches to topics requested by Fellows, including a research software engineering workshop to provide practical skills, and an interactive tutorial on variational inference led by Assistant Professor of Statistics and Data Science Aaron Schein.

Fellow-Led Programming

At the beginning of each academic year, Fellows may elect to serve on the Unseminar Committee, giving them room to shape their own training and develop the collaborative leadership skills they’ll need throughout their careers. The result is programming attuned to what Fellows actually need as they carry out AI-supported research. Beyond open problems sessions, the Unseminar has taken on a range of formats, with recent sessions including AI tutorials led by faculty experts, discussions of cutting-edge papers, research presentations by new Fellows, practice job talks by Fellows on the job market, a communication workshop led by by an expert in science communication, and planning sessions for the Annual AI + Science Hackathon and other programming led by the Fellows.

Konstantin Gerbig, a member of the fourth and newest cohort who researches planet formation in protoplanetary disks, chose to join the Unseminar Planning Committee when he began this fall. Soon, he and the committee began gathering input from the larger group about what they wanted from their weekly sessions and connected with the professional development committee to coordinate efforts.

“Coming in as a new Fellow, I appreciated that I could have a voice in shaping our experience from the start,” Gerbig said. “My main motivation for joining the Committee was that I felt that Unseminar would be very useful for myself and the other fellows. I very much subscribe to the idea of flexible programming, with room for informal interactions and community building, so I was excited to contribute to a successful execution of that vision. During my PhD, I co-ran a more traditional seminar series, so the Unseminar offered an opportunity to help organize something intentionally different.”

Preparing Future Leaders

When a conversation and a new skill can spark new lines of inquiry, the insights gained in Unseminar can shape research long after a Thursday afternoon session ends. Through these Thursday gatherings, Fellows build a network of peers who will serve as collaborators and thought partners throughout their careers. The Schmidt AI in Science Fellowship program prepares scholars for the boundary-crossing work of modern scientific discovery, and provides them with a forum to become the kind of scientists and leaders their interdisciplinary science needs.