Like so many great projects, the research that earned Ryan DeFever and Steven Hall three awards and a spot in a respected academic journal was fueled in part by coffee.  “Steven and I had probably 25 walks to Starbucks and back– at least–during the course of this research,” DeFever said. “There was a lot of chatting and bouncing ideas off each other.”

Sapna Sarupria (center) meets in her office with graduate students Ryan Defever and Steven Hall

DeFever and Hall were part of a team that had its research published this year in the journal Chemical Science. Their experience underscores how interdisciplinary research — sometimes enhanced by caffeine– energizes the educational experience that creates the next generation of engineers.

The team showed that the same machine-learning techniques that allow self-driving cars to see obstacles can be used to identify nano-sized structures of atoms and molecules, a tool that could help advance a wide range of research. Read the full article here.

The students involved in the recently published research were guided by two faculty members, Sapna Sarupria and Melissa Smith.   Dr. Sarupria, an Associate Professor in Chemical and Biomolecular Engineering, was the project leader.

Sarupria said that she was lucky to have DeFever and Hall working in her lab.   “Students, especially those who are creative and brave, are important to research,” she said.   “They do the work, but they’re also the ones who come up with ideas and motivate you,” Sarupria said. “These are my research collaborators– they are my true science collaborators, and they keep my energy going.”

Sarupria, DeFever and Hall collaborated on the research with Smith, associate professor of electrical and computer engineering, and her former student, Colin Targonski.    Smith said that she and Targonski contributed their machine-learning techniques and experience in applying them appropriately.   Just as important as the scientific discoveries, she said, is teaching students to collaborate across disciplines.

“I come from a national laboratory background where that is everyday practice,” said Smith, a former research associate at Oak Ridge National Laboratory. “That is why they make these big discoveries and extend science in big leaps and bounds. They work in an interdisciplinary team, rather than working with their own kind all the time.”

Targonski graduated in May with a master’s degree in computer engineering and now lives in New York, where he works as a machine-learning engineer at JP Morgan Chase & Co.

“This work was exciting because it offered an entirely new domain to work in– applying machine learning algorithms to molecular dynamics,” Targonski said of his work at Clemson. “We are especially excited about the state-of-the-art results we were able to achieve by using algorithms developed for the computer vision domain and adapting them to the computational chemistry domain.”

The research also helped DeFever earn a Ph.D. DeFever, who is from Greenville, graduated in August with his doctoral degree in chemical engineering.

When he crossed the stage at the hooding ceremony, he had two prominent awards under his belt, again thanks in part to the research. He received Clemson University’s Outstanding Graduate Researcher Award, and the Chemical Computing Group Excellence Award from the American Chemical Society’s Division of Computers in Chemistry.

DeFever is now considering whether to stay in academia as a post-doctoral researcher or pursue a job involving machine learning in industry or a national lab.

Hall, who is from Anderson, began the research as an undergraduate and is now a first-year Ph.D. student in chemical engineering. He also took home an honor based on the research, winning a best poster competition at July’s Rare Events Workshop at the Indian Institute of Science, Bengaluru.

Part of what made the research stand out is how fast it went. The team went from idea to published paper in about a year, with most of the work occurring in the final six months.

DeFever said one of his favorite parts of the research was that it involved lots of coding.  Most of the time, he said, was spent debugging code.

“The idea always is quick, but the implementation is long and tedious,” he said. “You get that moment where it all falls into place and it clicks and it works. And that brings you back for more because it’s thrilling when that happens. It’s worth a lot.”