This past week marks the end of my fourth semester at UVA, weeks after most students finished their last final exams. The reason for the prolonged sophomore year was the wrapping up of a two-year-long research project I joined when I first arrived at UVA. Working under the patient supervision of Josephine Lamp, computer science Ph.D. candidate and Jefferson Fellow, and in collaboration with UVA Health, this project has been the most fulfilling portion of my college education thus far.

The goal of the project was to develop a machine learning pipeline for the accurate prediction of heart failure. That pipeline can sit on a pulse oximeter device available for clinical usage as well as by out-patients. It’s a task that carries great weight: heart failure is the leading cause of death in the United States killing roughly 659,000 Americans each year. An at-home device that could predict heart failure with even moderate accuracy could save countless lives by getting at-risk people treatment before it’s too late.

After an arduous editing process, we finally submitted our paper with results for publication last week. This project was a series of firsts for me: first time conducting research, first time working with medical data, and first time writing and submitting a paper. Though it was certainly strenuous at times, I learned more from this project than I have from any of my classes, and more importantly, discovered a personal passion for research.

That’s not to say research is without its flaws. Projects all begin with a question; for us it was: can we predict heart failure using pulse wave velocity? Fortunately, that answer was yes. I spent countless nights tweaking our models to squeak out every remaining percentage of accuracy until we had strong results. But what if the answer was no? What if we exhausted all of our resources to discover that pulse wave velocity is not a predictor of heart failure? To me, that still seems like a worthy contribution. Researchers with similar questions would undoubtedly benefit from seeing our work yet no journal would publish those results. In that way, research feels overly competitive and not just an endeavor to further the academic pursuit for answers. That being said, I can appreciate the freedom to explore new ideas that research allows. Unbounded by a corporate mission and without a bureaucracy of command, I found peace in asking my own questions and trying my own solutions.

I’m certainly appreciative of the learning opportunity this project was. I developed technical skills wrangling irregular medical data and learning techniques for time series feature extraction and classification. The pipeline was developed for clinicians and patients not expected to have an understanding of machine learning and yet be able to deploy it effectively. This added the extra challenge of making the code user-friendly through careful instructions and helpful visualizations. Far beyond the technical skills, the project helped me to develop interpersonal abilities, collaborating and communicating with the many different people involved from fellow students to doctors at UVA Health. The long stretches of time where I had to work independently on large tasks forced me to improve my time management in order to meet strict deadlines. These are skills I certainly plan on drawing upon in future professional environments.

This project pushed me to my limits but has inspired me toward future research endeavors. In addition to the personal projects I will continue to embark on and showcase on this site, I’m looking forward to joining new cutting-edge collaborative research projects at UVA.