Roshini Balasubramanian: developing neural networks for medicine

Wednesday, Jul 7, 2021
by Sharon Adarlo

Roshini Balasubramanian, 21, Class of 2022

Studies:

Balasubramanian is majoring in operations research and financial engineering and is completing the undergraduate certificate program at the Center for Statistics and Machine Learning (CSML).

She is pursuing certificates in global health and health policy which is under the Princeton School of Public and International Affairs’ Center for Health and Wellbeing, cognitive science at the Program in Cognitive Science, and applications of computing at the Department of Computer Science.

Research:

In the summer after her freshman year, Balasubramanian interned at Children’s National Health System, a non-profit pediatric healthcare provider, through Princeton Internships in Civic Service. It was during this internship where she witnessed firsthand the transformational power of big data in healthcare. Healthcare practitioners were using data science to track patients and uncover new insights in other medical data they were collecting.

“This experience piqued my interest in data science as a whole and specifically artificial intelligence to improve patient care and efficiency in healthcare systems,” she said.

For her junior independent work, she followed this interest. She proposed a novel neurosymbolic combination of neural networks and encoded rules, and she demonstrated its applications in medical diagnosis. 

“Neural networks and deep learning have led to several significant milestones but require a large amount of training data, which may not be available in all cases,” she said. “Many models in healthcare are limited by data availability and demand high accuracy predictions for practical use.”

In response to these limitations in deep learning, researchers have proposed combining statistical AI methods like neural networks together with symbolic/rule-based AI techniques – the neurosymbolic method which Balasubramanian employed in her project.

“While these two methods have traditionally been separated, they share complementary strengths,” she said. “Bridging the gap between old and new schools of AI provides a powerful new approach to increase prediction accuracy and decrease dependency on large amounts of data.”

Her independent project for the CSML undergraduate certificate program will be part of her senior thesis and continues her interest in deploying machine learning techniques in medicine.

“My thesis will tentatively focus on statistical models of disease progression and applications for randomized controlled trials,” she said.

Balasubramanian found the CSML to be very helpful in increasing her skill set.

“My participation in the CSML program has largely shaped the courses I selected to take and my research interests. In addition to providing a guide for my major coursework, the CSML certificate program challenged me to take statistics classes outside of engineering like POL 346 – Applied Quantitative Analysis in the politics department and ones that went beyond the undergraduate level like ELE 571 – Deep Learning Networks,” she said. “Having taken a variety of classes through CSML, I feel confident that my data science skills will support my future endeavors.”

After graduation, Balasubramanian is interested in joining industry with the goal of solving healthcare challenges with innovation

“This summer, I am working in biotech investing at a hedge fund,” she said about her work at the firm, Point72.

“Long-term, I plan to leverage my background in machine learning and interdisciplinary problem solving in my daily work,” she said.

Her other work experiences include stints as a fellow at ScaleHealth, a healthcare innovation consortium; product manager at startup Health AI; market research summer analyst at Orange Vista, a consulting firm; and intern at the financial engineering division at Fannie Mae.

Extracurricular Activities:

Balasubramanian is an oboist in the Princeton University Orchestra and the chamber group, Princeton Camerata. She is also co-editor-in-chief of the Princeton Public Health Review, co-founder of the Princeton Students vs. Pandemics chapter, a peer academic adviser, and a peer health adviser.

For Fun:

Balasubramanian enjoys listening to podcasts, trying new foods, and hiking in her free time.