Balancing Baby Bottles and Biometrics is Ray Carpenter’s Formula for a Promising Future

By Dave DeFusco
Ray Carpenter’s journey to becoming a Ph.D. student in materials science in the Department of Applied Physical Sciences at UNC is as much a testament to his intellectual curiosity as it is to his ability to balance family life and his academic pursuits. In his first year at the age of 30, Carpenter has immersed himself in computational neuroscience, a field that blends his skills in data science and machine learning with his passion for solving complex problems, but his path to academia has been anything but linear.
Carpenter and his wife, who were high school sweethearts, have pursued their academic and professional careers in lockstep. Both graduated in May 2024—she with a master’s degree in cybersecurity after nine years in the military and he with a bachelor’s degree in economics and a minor in data science from UNC. As if their lives weren’t already full, they are parents to three children, ages 9, 4 and 1.
“Being a dad is fun. There’s never, ever a dull moment,” said Carpenter. “My children want my attention all the time, and I enjoy being with them. Parenting takes a lot of teamwork. I spend two afternoons at home a week, and I work from home another two or three afternoons, depending on the week. Those days, I’ll take care of them and then work late at night after they’re asleep.”
The decision to pursue a Ph.D. now was a strategic one for he and his wife. “It’s better to rip the Band-Aid off now while the kids are young,” he said. “Since my wife and I are both 30, we decided to get it over with. I really enjoy the work, and I’d rather make more money later and for a longer period of time.”
Carpenter has extensive experience in computational research methods, machine learning and data analysis. During his undergraduate years, he conducted computational modeling in health economics, which laid the groundwork for his current research. Later, he worked for a year and a half in the Department of Statistics and Operations Research at UNC, analyzing single-neuron firing patterns from NeuroPixel data to predict visual stimuli in mice.
“That experience created a direct path from economics to materials science,” he said. “We used single-cell models to understand neuronal encoding. It was fascinating to see how computational techniques could predict what a mouse was looking at based solely on neuronal firing patterns.”
In his Ph.D. program, Carpenter is co-advised by Professors Nicolas Pégard of APS and Jose Rodríguez-Romaguera of the UNC School of Medicine, whose expertise spans materials science, genetics and neuroscience. In collaboration with his advisors and Dr. Rebecca Grzadzinski, a clinician-scientist at the UNC School of Medicine, Carpenter is studying the link between pupillometry biometrics and behavioral phenotypes in infants at risk to develop neurodevelopmental disorders. By studying how pupillary responses correlate with brain states, Carpenter aims to develop disease classification algorithms for conditions like autism spectrum disorder (ASD) in infants from 6 months to 2 years old.
“At that age, infants can’t tell you how they feel or act,” said Carpenter. “We’re trying to use pupillometry to figure out an average pupillary response to social stimuli and see if it differs between ASD and non-ASD individuals. Once we’re able to establish these baseline responses, we can use them for other disease classification. Since this methodology isn’t limited to autism; it could be applied to any social disorder where social dynamics are a factor.”
Carpenter’s work builds on his previous research with the National Science Foundation’s Research and Training on Networks program in Chapel Hill. There, he benchmarked machine learning and deep learning models to predict visual stimuli in mice. Among the models tested, Long Short-Term Memory (LSTM) networks achieved the highest prediction accuracy of 96.6%. Other related research involves examining factors contributing to cognitive decline in mid-to-late-life populations and forecasting life expectancy for over 40,000 individuals.
“Our study sought to identify machine learning models capable of predicting visual stimuli based on single-neuron spike patterns,” he said. “Drawing insights from neuronal recordings is integral for understanding the neural mechanisms underlying cognitive processes. These models could aid in developing brain-computer interfaces and disease classification tools.”
Carpenter credits much of his success to the support of his advisors. “Nico and Jose trust in my skills a lot, and they’re the best advisors I’ve ever had,” he said. “They’re incredibly understanding about my need to balance family life with work and research. They’re flexible about when I work, as long as I put in the time, which has been invaluable.”
January 20, 2025