CSpotlight: At home with computer science research
As a natural problem solver, new B.A. graduate Gabe Mersy discovered that computer science offered him the perfect blend of mathematics, statistics, and coding. Through multiple on-campus research opportunities and internships, he determined his future was in research (versus industry), and this fall, will start his studies towards a Ph.D. in computer science at the University of Chicago.
Why did you choose to study computer science at the University of Minnesota?
Having grown up in Minneapolis, I have always considered the University of Minnesota to be a top-tier university. The computer science department is known for its cutting-edge research in addition to its outstanding undergraduate program. I took most of my core classes with exceptional teaching lecturers and most of my electives with field-leading researchers. This has allowed me to develop solid fundamentals alongside an ability to solve problems that haven’t been solved before.
Plus, the state of Minnesota and the city of Minneapolis are both great places to live! I enjoy the amazing park system, the bodies of water, and the overall size of the Twin Cities. The campus is also great, and there are ample opportunities to find your niche.
What sparked your interest in computer science? How does it relate to your interests in statistics and applied mathematics?
I entered my first semester as a statistics major. After taking my first computer science class, I realized that I was more drawn to the sort of problem-solving where you end up implementing solutions in code as opposed to just writing them on paper. I found that computer science contained a perfect blend of mathematics, statistics, and coding. Machine learning and classical artificial intelligence both rely heavily on statistics and probability respectively, so I am still able to pursue my interests in data science while studying computer science.
Tell us more about your experience being an undergraduate research assistant.
I have worked on a large number of research projects spanning quite a few disciplines. My first research assistant experience was under Professor Glen Meeden (School of Statistics), in which I developed some computational tools for his work on Bayesian sampling theory. After that, I worked with Professor Richard Landers and his Industrial-Organizational Psychology lab to develop data and feature engineering pipelines for an applied machine learning project.
I also collaborated with one of my peers, Jin Hong Kuan, on two research projects related to deep learning. As a couple of allegedly nerdy music producers who are interested in AI, the first project was centered around improving deep learning-based approaches to music genre classification. We ended up publishing our work as an extended abstract at the 2021 Association for the Advancement of Artificial Intelligence (AAAI) conference. The second project had to do with tackling the problem of model bias that can manifest as accuracy disparities between certain groups of people. This sort of model bias tends to reinforce real-world social bias. We proposed a new way of addressing this problem at the model, as opposed to dataset, level and obtained some promising preliminary results.
What did you find to be the most challenging and most rewarding aspects of conducting research?
Research is challenging because it deals with very open-ended questions that often have no clear answers. However, in my opinion, there is no better feeling than achieving a promising result after countless failures!
Throughout my undergraduate experience, I have been especially fortunate to have been mentored by Maria Gini. Professor Gini has provided me with instrumental feedback that has dramatically improved the quality of my work and has ultimately helped me develop as a computer science researcher. Her long-standing efforts to both inspire and support the future of computer science is a testament to the department’s commitment to broadening participation in computing.
I also had the privilege of attending the AAAI Undergraduate Consortium, where I received invaluable advice on applying to grad school and advice about pursuing an AI research career. I presented my contributions to an AI research project and learned about the work of phenomenal undergraduate scientists from across the world.
What were some of the highlights of your internships?
I completed two summer internships: one in data science and the other in software engineering. My data science internship was at a political data analytics company called The Burr Project, where I worked on an applied machine learning research project concerning the prediction of political polarization from demographic data. The team consisted of my supervisor and a handful of undergraduates from different colleges. We produced a publication that I presented at an IEEE computer science conference last summer. My software engineering internship was at U.S. Bank. I worked on features for a large-scale payment processing system and gained important industry skills such as agile and full-stack development. Both of these internships were memorable in their own ways, but I did learn that I tend to be more drawn to research than industry software engineering.
Tell us about your work with Kite. How did you become involved with their YouTube channel?
Kite is an AI-based code autocomplete company. When I started there, I was originally working on a large collection of Python documentation that is used as a sort of automatic reference that pops up when you write lines of code that use common libraries. Eventually, that project ended, and they started a marketing campaign on YouTube. I wrote scripts and code for various YouTube videos surrounding interesting topics in AI. For one video, I used a pre-trained convolutional neural network to classify the emotions present in audio clips of Michael and Dwight from the popular TV show The Office. In another video, I developed a recurrent neural network for predicting Tesla’s stock price.
What advice do you have for incoming computer science students?
I would encourage students to try to discover the applications of computer science that fit their passions. Technology is unique in that it is ubiquitous, so as a student, you are developing tools that quite literally shape the virtual world on a very large scale. With this comes a great deal of responsibility. It is very important to consider the ethical and societal impact of technology. Interesting technical problems always benefit from a human-centered way of thinking.
Also, don’t be afraid to go outside of your comfort zone. Take classes that push you to think outside the box. Almost every field has some degree of intersection with computer science. As a computer scientist, you have the opportunity to take a formerly abstract idea and implement it in software. There’s truly nothing more rewarding than coding up something that you are truly passionate about.
What are your plans after graduation?
I am going to be starting my computer science Ph.D. at the University of Chicago in the fall.