While a junior in high school, Grayson Hoy ’23 could not wait to get into college and finally be able to do research.
“You can actually do research now,” said his brother, who was a William & Mary student at the time. This led Hoy to reach out to the late Professor Samuel Abrash from the University of Richmond, in whose chemistry lab he interned during the summer between his junior and senior years of high school.
Once at William & Mary, Hoy soon started working in the research group led by Kristin Wustholz, Mansfield Associate Professor of Chemistry. Since then, he has never stopped pushing the boundaries of chemistry and machine learning. Recently, he secured an award from the National Science Foundation Graduate Research Fellowship Program, which will support his graduate research in computational chemistry at Yale University.
A senior majoring in chemistry, Hoy is one of just over 60,000 fellows who since 1952 have been identified by NSF as researchers likely to contribute to the scientific and technological advancement of the nation.
As the United States’ oldest program of its kind, the NSF GRFP funds outstanding students pursuing research-based graduate careers in STEM at accredited U.S. institutions. The fellowship, which is valid for five years, provides three years of financial support, including an annual stipend and a cost-of-education allowance.
The program’s notable alumni include Sergey Brin, Google co-founder; Steven Chu, former secretary of energy and 1997 Nobel Prize in Physics; as well as 41 other Nobel laureates.
A ‘T-shaped thinker’
Despite Hoy’s early commitment to research, it was not obvious he would continue on a research path. He had enrolled into William & Mary thinking he would be a pre-med student, but his mind kept going back to his time in the lab.
“I think that what helped me realize I wanted to do research was coming in every day excited to do the work,” remembered Hoy.
“But I don’t want to scare people and have them think, ‘If I want to be a scientist, I have to be excited all the time; I always have to enjoy my work,’” he added quickly. “This is not true, there are ebbs and flows, and that’s OK. That’s part of life. But there is this idea that you want to keep coming back to science.”
“Four years ago, when I expressed my interest in doing chemistry and data science, it was not something so common at the time,” he continued. “Professor Wustholz really embraced that and let me explore both my interests at the same time, which made me a very T-shaped thinker – expert in one field but with a good level of understanding in others.”
In the Wustholz group, Hoy supported the development of a novel imaging technique based on the blinking behavior of molecules — that is, the patterns in which they emit light.
“Nanoscopic reporters are used to report, through emitting light, what they observe in their surroundings,” said Hoy. “Scientists often use the color of the emitted light, but this can really limit the types of reporters you use: For example, two reporters can indicate opposite diagnoses, but they both emit red light so you can’t tell them apart.”
Hoy developed an automatic way of identifying the reporters using machine learning, serving as a liaison between chemists in the Wustholz group and the undergraduate data scientists who had been involved in the project.
By implementing a deep learning model, what had previously required one month’s worth of work by three students was possible in five hours and with the same accuracy.
Hoy’s third paper — and his first as lead author — took this concept even further.
But it was right indeed, and the model obtained allowed a rapid, generalizable and accurate classification method, opening to new opportunities in single-molecule imaging.
Everything everywhere all at once
“We can’t look at all data at once; it’s too much for us as humans to understand and process. With machines, this is what they’re really good at,” said Hoy.
He cited solar energy and pharmaceuticals as two of the areas where the intersection of chemistry and machine learning can produce very tangible results.
Dye-sensitized solar cells can generate renewable energy at a lower cost compared to traditional silicon-based cells. These third-generation cells are cheaper and easier to manufacture, since they use readily available materials such as organic dyes and titanium dioxide. However, they do not currently convert energy as efficiently, and experimentally searching for organic dye sensitizers that are more viable than others is expensive and time-consuming.
“We currently use certain structural motifs to find viable dye sensitizers because it’s what we gathered parsing through data from human experimentation. But a generative machine learning model doesn’t have to rely on these motifs,” said Hoy.
Machine learning is already being used in the pharmaceutical industry; however, advances in deep learning models promise to further accelerate drug discovery. Thinking about his future studies, Hoy realized that studying just a few molecules could take an entire Ph.D.’s worth of time. “But if you’re able to train a model in a month, then it could help you find five very viable candidates during your Ph.D.,” he added.
Hoy’s enthusiasm is indicative of how much machine learning can keep enhancing chemistry. At the same time, his attitude highlights persistence as a fundamentally human component of research.
“It really excites me that you can go in and learn about these different techniques to accomplish certain goals you come up with, figure out the questions you want to ask and once you learn these techniques, you can answer these questions, and they may be questions that had not even been asked before.”
With support and mentorship
A Ph.D. in chemistry may lead to a career in the industry, in a national laboratory or in academia.
Regardless of what the future will bring, Hoy knows that he will devote part of his time to training and mentoring people, “especially those who face extraneous barriers when pursuing science.”
“I recognize I come from a position of socioeconomic privilege and that I could pursue opportunities that may not be as easy to access,” said Hoy, who aims to be instrumental in helping remove barriers and increasing access and representation.
Hoy observed that many people also become discouraged because they don’t see results immediately, and may abandon science if they don’t get support. “But the whole thing about science is failing and failing until you succeed; and I want to allow these people to see themselves in science, bring them into a chemistry lab; and, if possible, give them a day when they can come into the lab and do something,” he said.
Hoy has already been serving in several mentoring and supporting roles, including a chemistry teaching assistant, an orientation aide and a research ambassador for the Charles Center for Academic Excellence. Having — remarkably — received his graduate fellowship while still an undergraduate, he now urges his peers considering graduate school to be on the lookout for funding opportunities.
Thinking about research environments, Hoy cited the “growth mindset” he perceived as typical of William & Mary, “the idea that you can always grow and become a better scientist through work and support,” he said.
“Research has played a very big part in transforming how I think and in determining my career path,” said Hoy. “I’ve got to this point, in part, because of all the work I’ve put in, but also thanks to the support I received from family and friends and mentors like Professor Wustholz, my lab mentors Kelly Kopera and John Li and Professor Abrash, who sadly passed away this past February.”
Editor’s note: Data is one of four cornerstone initiatives in W&M’s Vision 2026 strategic plan. Visit the Vision 2026 website to learn more.
Antonella Di Marzio, Senior Research Writer