Following a year-long collaboration, a Trinity College student and faculty member recently published their research in a peer-reviewed scientific journal. The project by Shivanshu Dwivedi ’26 and Senior Lecturer in Physics Kalum Palandage applies machine learning to streamline complex many-body calculations.
The student and mentor met when Dwivedi was in his first year at Trinity, but they first began discussing this project prior to the summer of 2025. The project was made possible through the support of the Trinity College Summer Research Program.
Their work, which began as an early conversation about extending an existing theoretical model, eventually developed into a paper. “Deep Reinforcement Learning for Autonomous Control of Hole‑Doped Hubbard Clusters: A Comparative Study” was published in Applied Physics Letters (APL) Machine Learning (Vol. 4, Issue 1) in March 2026. According to the paper’s abstract, “This work establishes autonomous reinforcement learning (RL) as a viable, highly efficient framework for rapid optimization and non-obvious strategy discovery in complex quantum systems.”
Below, Dwivedi and Palandage reflect on the process of conducting research and writing a paper together.
Dwivedi: I’ll give you a simple analogy: Let’s suppose you wanted to buy a specific ice cream and there are 1,000 stores in front of you. And let’s say you want to buy the ice cream wherever it’s cheapest. So what we were doing initially was going to every store and trying to note the price of ice cream at each one. Professor Palandage and I came up with a new method in order to address this problem because our computers weren’t able to handle all these heavy calculations. We used machine learning, which will try to make predictions of which store will have the cheapest price. So instead of visiting 1,000 stores, we just have to analyze 10 stores in which the price could be cheapest. Then, we just visit those 10 stores and try to find three stores where it’s cheapest, and then find the cheapest one. This reduces the competition from 1,000 to four of five.
Palandage: Our goal is to control interacting electrons using external energy inputs, such as an applied electric field. In the Hubbard model, the standard approach involves calculating energy levels for every possible configuration, which quickly becomes extremely complex. With this new method, we can avoid that exhaustive process while still capturing the system’s behavior efficiently.
Palandage: Shivanshu is kind of an expert in machine learning. We had weekly meetings that initially started by making sure Shivanshu understood the model, addressing what our limitations were at that point, and then setting our goals. As we continued working, our meetings were meant to assess our progress and, sometimes, to cut down on goals by deciding what we wanted to focus on.
Dwivedi: From the student side, those weekly meetings were essential. We sat and discussed the way we should approach a specific problem, or maybe to not do this and try this new thing. I might have gotten lost if Professor Palandage didn’t guide me in the right direction. But in addition to that, Professor Palandage gave me a lot of liberty to try new approaches and experiment to see what might give us better results.
Palandage: After collecting a strong set of results, we worked through multiple revisions to determine the most effective way to tell the story of our findings. In the following weeks, we searched for an appropriate journal and ultimately decided to submit our work to Applied Physics Letters.
Palandage: Even if you do not publish a paper, getting the experience of undergraduate research is very important. You learn how to manage your time with your coursework and your research, and I think that’s a valuable experience for undergraduate students. There are many ways to take advantage of it: You can spend the summer on campus for that experience and work with other students, or you can go to research conferences to present your work and get some experience on presenting.
Dwivedi: I completely agree. Being able to do research as an undergraduate is a huge privilege that you get at colleges like Trinity, where you can work directly with the professor and where your ideas are really valued. That’s not something you get everywhere. At a lot of other institutions, that might not happen because there are so many other people.
Dwivedi: I’m really grateful to have such a helpful mentor, not just in research but also for advice about how I should pursue my grad school application or how I should plan for my future in the industry as a software engineer. The experience of working with Professor Palandage and other professors in the Physics Department exposed me to how beautiful the actual science part of physics is.
Palandage: I was really very lucky to have Shivanshu as my research student. He’s very talented and has a great future ahead of him.
Two Trinity College student-led teams have been selected to receive Projects for Peace grants for summer 2026, joining a long tradition of Trinity students designing community-based solutions for local challenges worldwide.
Trinity College’s Volunteer Income Tax Assistance (VITA) clinic is wrapping up its most productive year to date, with a goal to have students prepare at least 760 free tax returns for Hartford residents by the April 15 filing deadline.
For 20 years, the Trinity International Hip Hop Festival has brought together artists, students, and community members from Hartford and across the globe. A new photo exhibit opening in conjunction with this year’s festival celebrates its history and legacy.
A traveling environmental justice exhibition created by Trinity College students, faculty members, and community partners will be featured in Trinity’s Mather Hall Art Gallery from April 1 to 29, 2026.
