Naveen Raman and Naveen Durvasula Named 2021 Goldwater Scholars

Naveen Raman, a junior computer science and mathematics double major at the University of Maryland, and Naveen Durvasula, a sophomore dual-degree student in electrical engineering and computer science and business administration at UC Berkeley who has been conducting computer science research at UMD for five years, were awarded scholarships by the Barry Goldwater Scholarship and Excellence in Education Foundation. The foundation encourages students to pursue advanced study and research careers in the sciences, engineering and mathematics.

Over the last decade, UMD’s nominations yielded 37 scholarships—the second most in the nation behind Stanford University—and a UMD computer science major has been named Goldwater Scholar every year since 2016. 

While Raman and Durvasula share the distinction of being named Goldwater Scholars, they also share two mentors in UMD’s Department of Computer Science: Distinguished University Professor Aravind Srinivasan and Assistant Professor John Dickerson. 

Durvasula, who grew up in Potomac, Maryland, began working with UMD faculty members in 2015 as a middle-school student, and Derwood, Maryland-native Raman began working with them in 2018 as a senior in high school.

Since then, Raman—who is also a member of the Advanced Cybersecurity Experience for Students in the Honors College—has published four papers and submitted a fifth for publication.

He began by developing algorithms to identify cancer mutation signatures with Srinivasan and former Assistant Professor Max Leiserson and moved on to working with Dickerson to develop policies that balance fairness and profit in ride-pooling systems.

He’s also currently working with Associate Professor Jordan Boyd-Graber to improve question answering systems by leveraging data from trivia competitions. Raman’s focus is on advancing so-called named entity linking algorithms, which connect names found in a question to larger repositories of data about them like Wikipedia. These advances will ultimately help question answering systems perform better on a diverse set of questions.

“Naveen Raman is a clear star researcher—and practitioner—in the making,” Dickerson said. “He is driven, questioning, curious and technically talented, as well as a young adult with a strong sense of civic duty and commitment to using technology for social good.”

In Summer 2019, Raman worked to detect rudeness, toxicity and burnout in open-source communities as a participant in Carnegie Mellon University’s Research Experience for Undergraduates in Software Engineering program. Last summer, he worked at Facebook to develop a user interface for debugging machine learning models and learned about important societal issues that machine learning can help solve, such as hate speech detection.

An active competitor, Raman’s team won the National Academy Quiz Tournaments’ Division 2 Intercollegiate Championship Tournament during his freshman year. In 2020, he and two classmates received an honorable mention award in the 72-hour Mathematical Contest in Modeling for their project that analyzed the effect that rising global temperatures have on herring and mackerel fishing along the Scottish coast. He also received an outstanding award in the 2020 SIMIODE Challenge Using Differential Equations Modeling for his team’s work on modeling interactions in refugee camps.

He has been a teaching assistant for a programming languages class and the lead student instructor for a class on algorithms for coding interviews. He also serves as vice president of UMD’s Puzzle Club.

Off campus, Raman teaches literacy skills to underprivileged elementary school students in the Maryland Mentor Program and volunteers at the College Park Academy charter school helping students improve their math skills.

He has been awarded the President’s Scholarship, Brendan Iribe Endowed Scholarship, Capital One Bank Dean’s Scholarship in Computer Science and Corporate Partners in Computing Scholarship.

Raman plans to attend graduate school to pursue a Ph.D. in computer science, with a focus on the fairness of artificial intelligence algorithms in critical fields such as criminal justice, job markets and health care.

Durvasula is interested in problems that lie at the intersection of theoretical computer science, machine learning and mechanism design. He has co-authored two published papers and submitted two more for publication on his research at Maryland.

He began conducting research at UMD with Srinivasan and Dickerson on a mix of theoretical and applied problems in optimization and learning.

“Even though I was in middle school, Aravind treated me as he would any other student, encouraging me to engage with graduate material on randomized algorithms,” Durvasula said. “I am grateful to him and John for helping me realize the joy of research and the research process.”

The trio began working together on a project to improve kidney exchange programs. These programs have saved thousands of lives by matching people who need kidneys with willing donors—typically total strangers. But the system is challenging. 

Not all patients are equally easy to match, and donor organs are not of equal quality—factors that affect medical and insurance decisions. Some patients are matched within weeks, while others may wait for years with no match offers at all. 

To help address these challenges, Durvasula developed a method to estimate organ quality and waiting time for a kidney for a specific patient and paired willing donor who enter an exchange. He has verified the accuracy of his model on real U.S. kidney exchange data using the most accurate kidney exchange simulator.

In the future, the researchers see this model being used a decision-support system to help inform doctors, patients with end-stage renal disease and would-be kidney donors about critical planning decisions—such as accepting a "weaker" proffered kidney now instead of waiting for a "stronger" potentially available kidney in the future.

“Naveen Durvasula has the unique ability to quickly and deeply understand a diverse array of topics in pure mathematics, statistics and computer science. He is also very clearly driven by a desire to see both near- and long-term impact,” Dickerson said. “Naveen is also considerate, giving—and forgiving, when I ask him for the third time to explain an idea more slowly—and he has a research and work ethic that I would pay to instill in all students.”

Durvasula is also working with Srinivasan and Dickerson on a project in the area of reinforcement learning, a branch of machine learning that aims to solve problems related to decision-making. 

While modern algorithms can solve a variety of different tasks, they need to train by repeating the task thousands of times. However, certain tasks, such as negotiations, can only be attempted once. Durvasula developed an algorithm that succeeds in this severely constrained setting by creating successively better approximations of the environment that can be repeatedly trained on.

From 2016-18, he worked with UMD Computer Science Professor William Gasarch on the so-called “Muffin Problem.” This problem asks: What is the best way to divide up m muffins for s students so that everyone gets m/s muffins, with the smallest pieces maximized?

In addition to being invited to the 2018 Research Science Institute, Durvasula served as head teaching assistant for a six-week statistics course last summer for the Summer Stem Institute. He also co-founded Aerostry, which is developing a balloon-based system to help firefighters detect and monitor wildfires more effectively. 

Durvasula plans to apply to Ph.D. programs in computer science and specialize in theory and artificial intelligence.

“It is such a pleasure to work with such highly motivated students,” Srinivasan said. “I look forward eagerly to the ways in which Naveen Raman and Naveen Durvasula will impact our world positively and significantly.”

Written by Abby Robinson