Connecting Dots: Meet Angikar Ghosal, the Duke Senior Writing 4 Honors Theses

October 26, 2023
Science Magazine

Above: Angikar Ghosal. Image courtesy of Angikar Ghosal.

How does one student end up pursuing research projects on topics ranging from error correcting codes to theoretical neuroscience to political theory, leading to four separate graduation with distinction theses? 

For Angikar Ghosal (Trinity ’24), the answer lies in a little bit of rebellious nature, taking six classes a semester, and an intense drive to forge interdisciplinary connections.

Ghosal, a native of Kolkata, India, was a pre-engineering student in his country’s “infamous” secondary school system. For students at the highest level of the pre-engineering track, high school means a grueling four years of math, physics, and chemistry in preparation for the Joint Entrance Exams (JEEs) and admission to the Indian Institutes of Technology (IITs), the nation’s premier engineering schools.

The JEEs are one of the most challenging college entrance exams in the world. Performing well among millions of students on the JEEs is the only way to earn acceptance into an IIT, a beacon of social mobility.

Students, including Ghosal, often enroll in cram schools and are told to focus solely on exam preparation, often at the expense of other academic interests. Although Ghosal enjoyed math and physics, he refused to solely focus on the JEEs. 

Instead, he “rebelled” by self-learning a whole host of other subjects: philosophy, astrophysics, and computing. Linguistics, however, was Ghosal’s “original love.” He represented India at the International Linguistics Olympiad four times, becoming the country’s first gold medalist.

“It was the way I rebelled against physics [and] math when I did not want to do physics,” he said. “When teenagers rebel, they get girlfriends and stuff, right? I did linguistics instead.”

But by the time he’d graduated high school, Ghosal had done it all. His all-India rank was 161 on the JEE Advanced, which was more than enough to earn admission to top IITs for computer science. And, his rebellion was successful. He had studied a gamut of disciplines, setting the stage for his time at Duke.

A semi-random walk through Duke

In the fall of 2019, Ghosal arrived at Duke, giving up his IIT spot for a U.S. university. Why? Because the U.S. system gave him the freedom to explore, he said.

Ghosal maintained his interest in math and computer science, but he became interested in the intersection of neuroscience and economics, as well as traditional bench sciences. Throughout his first year, he took introductory and intermediate classes in many of these disciplines while working in a cognitive neuroscience wet lab.

But towards the end of his second semester on campus, the COVID-19 pandemic hit. Labs across the University closed, including Ghosal’s, and his interests evolved as a result. At this point, Ghosal had started taking six classes a semester.

“If COVID-19 did not happen, I would not have left wet lab research that early,” he said. “I was interested in both quantitative and natural sciences. But COVID-19 was a big reason why I shelved the plans for natural sciences a bit earlier than I would have.”

During the 2020-21 school year, when he stayed on Duke’s campus as an international student, Ghosal worked for Professor of Computer Science Cynthia Rudin. He studied causal inference and completed his first natural language processing project, both of which led to peer-reviewed publications. This was the first time he saw how linguistics and the quantitative sciences interacted with each other.

As a sophomore, Ghosal would also see how the natural sciences coincided with linguistics through a graduate-level organic chemistry class he took in the fall.

“The reason I loved organic chemistry is because I saw that [it] is this hieroglyphic-like language with its own grammar. For me, organic chemistry was like a language.”

In the spring of 2021, he saw how the natural and quantitative sciences could intersect by taking a graduate class in theoretical neuroscience. Ghosal wanted to learn about neural networks and artificial intelligence “from the actual roots.”

“[The course was] basically applications of every field of math that you can imagine, but to neuroscience problems,” Ghosal said. It was through this course that he realized that math provides the “abstraction” that is needed for “understanding phenomena that are very non-mathematical.”

Following a gap semester that fall, Ghosal had a flurry of formative research experiences and courses that built upon everything he’d learned and solidified his academic future at Duke. 

After returning, he began a project with Nicolas Brunel, distinguished professor of neuroscience at Duke School of Medicine, that has now morphed into one of his theses. A little bit of luck led Ghosal to another natural language processing project, this time with Bhuwan Dhingra, assistant professor of computer science.

Dhingra followed the beaten path that Ghosal didn’t take, going to an IIT to study engineering, and Ghosal wanted to learn more about Dhingra’s journey to academia. 

Dhingra ended up teaching Ghosal natural language processing from the very beginning. For six months, Ghosal read papers, getting up to speed with the literature. Now, he runs experiments that form the crux of his second thesis. 

“The experiments continue. There have been some setbacks, simply because of the fact that … I'm one person trying to research in NLP,” he said. “And I have six mega-corporations doing research at a faster rate than I am, and I can't control that.”

In the fall of 2022, Ghosal took MATH 501, Algebraic Structures I, with Robert Calderbank, former Bell Laboratories vice president for research. This course “arguably changed my life,” he said. Not only did it cover abstract algebra, the course also contained content inspired by Calderbank’s industry experience.

The 20-30 hour weekly problem sets showed Ghosal that abstract algebra is foundational to not only research in applied fields like communications and quantum computing but also other mathematical disciplines like number theory and combinatorics. 

“[MATH 501] was the culmination of the philosophical attitude I had towards science,” Ghosal said. “When you're good at code, when you're good at computation, and with mathematical reasoning and abstraction, you can pick up domain knowledge in lots of different fields.”

Along the way, Calderbank became a close mentor for Ghosal. Now, Ghosal is working on an applied math thesis with him on quantum error correcting codes, his third thesis project.

Ghosal decided to write his fourth thesis project in the fall of 2023. The project ties his quantitative linguistic knowledge to his coursework in the humanities. Working with Professor of Political Science Michael Munger, he’s studying how predictability of socioeconomic discourse translates to indifference and cynicism—not from a data science or computational perspective, but through a literary and philosophical lens.

‘Every single field’

Ghosal is majoring in math and computer science and minoring in economics. He will graduate Phi Beta Kappa from Duke and is currently in the process of applying to doctoral programs, but he remains undecided on his field of study.

“Right now, I see myself as an applied mathematician and a computer scientist, who knows math and knows coding,” Ghosal said. “I want to apply them in every single field possible.”

He enjoys teaching, having served as a grader or TA for over 20 classes, and wants to stay in academia. The same attitude that prompted him to study linguistics back in high school, along with the mathematical acumen he developed at Duke, is what will drive his future career.

“In many ways, fields of study are like their own languages with their own universalisms, and those universalisms are ways you can approach learning a new subject,” Ghosal said. “... It's the urge to discover the systemic in certain fields and try to find structures which you think must exist, to find structures in existing bodies of knowledge.”

“And that’s where you combine the power of abstraction that you learn from mathematics with the linguistic urge to get to know the field better.”

But for two more semesters, Ghosal is doing what he loves most: sitting in on classes that interest him, conducting research, and connecting dots.

Related Articles