Primus Opus

In high school, I published my first paper. After listening to an NPR story about government agencies suspending pollution regulations during the coronavirus pandemic, I submitted a research proposal investigating this topic to Harvard College's Coronavirus Visualization Team and became the project lead of an international team there. Through a comparative case study, we discovered that the regulation suspension during the pandemic could result in higher levels of pollution. While companies hoped that suspending these regulations would result in improved financial health, this goal was not met as unemployment levels continued to skyrocket during the pandemic. We published a paper in Environmental Sciences Europe urging governments to repeal suspensions and reinstate prior pollution regulations.

This project was my first exposure to research, machine learning, statistical analysis, and policy. It was this exposure that inspired me to further research at the National Science Foundation. Because machine learning was the area that I was the most unfamiliar with, I sought to study the area as a government intern.

University Research

Carnegie Mellon University's Research Experience for Undergraduates in Software Engineering (REUSE) marked my first research experience with a university. I worked on research studying broadband during the pandemic and diversity within the cybersecurity field. Using quantitative data collected by Microsoft and Federal Communications Commission and interviews conducted at cybersecurity conferences, I drafted proposals concerning cybersecurity education. Such proposals detailed how cable could accelerate broadband access in America as well as how early support can help make cybersecurity more inclusive for women. I presented this work to Egypt's Board of Education as well as the National Security Agency.

I attended this program the summer after my first year at UVA. REUSE was what convinced me to major in computer science and attend graduate school because my mentors demonstrated how computer science could incorporate other fields. I learned how to perform open coding--a qualitative analysis technique--on interviews. I tagged transcripts to find commonalities between the anecdotes of multiple female engineers. While this method has roots in the medical field and psychology, I use open coding in natural language processing research. With open coding, I can assess how closely a language model aligns with human thought.

Government Research

The following summer I worked at the NSF's Technology, Innovation, and Partnership Directorate and collaborated with the agency's Data Analytics Group. Because the CHIPS and Science Act provided 81 billion dollars in funding to the NSF, I was tasked with determining whether the agency was on track to meet the goals delineated in the piece of legislation. Additionally, I determined whether large language models could be used to automate government work. Using qualitative analysis, I discovered that the NSF was funding all projects laid out in the CHIPS and Science Act with an over-representation of artificial intelligence research. I also discovered that large language models were generalists and required finetuning to acknowledge the agency's unique needs. Moreover, I trained smaller models that were more accurate than the large language models in processing agency documents. I presented these findings to Assistant Director Erwin Gianchandani. My time at the NSF inspired me to pursue natural language processing for graduate school.