DIAMONDS program accelerates undergraduate data science research

Thanks to the DIAMONDS (Directed, Intensive and Mentored Opportunities in Data Science) program, undergraduate students from across the country spent ten weeks doing intensive data science research on the Tufts campus this summer. DIAMONDS is open to all undergraduate students across the United States and each year the cohort features a mix of students from Tufts and other universities. In the program’s fourth year, a dozen students from all backgrounds were accepted, including five from Tufts University.

Each student receives mentorship from a Tufts faculty member who oversees their work on a data science project in their lab. Luciano Galvani, E26, worked with Ada Lovelace Associate Professor Hari Sundar and PhD candidate David Van Komen of the University of Utah. His project focused on making accelerated large scale symbolic computation more efficient. Scientists use these computations in physics simulations such as black hole research. They typically consume a lot of time and computer processing power, but Galvani helped develop an algorithm that streamlines the process by tracking known dependencies to keep traversals in a logical order.

Another group worked to improve algorithms that prioritize candidate disease genes, based on network data that captures their interactions. Testing these genes could facilitate the discovery of new drug targets for conditions such as Alzheimer’s disease, arthritis, asthma, and more. Working with Professor Lenore Cowen, computer science major Amelie Hopke, A27, and undergraduate student David Yee of the University of Denver tested several adjustments to the ADAGIO algorithm to give more consistent and accurate output across a range of diseases. Although ADAGIO is already high performing compared to other methods, their work investigated additional ways to fine tune the algorithm.

From 2021-2024, DIAMONDS operated as a National Science Foundation Research Experiences for Undergraduates (NSF REU) site. This year the program’s three-year grant expired, but the program was able to continue due to leftover funds from the NSF REU grant, faculty mentor grants, and a handful of private anonymous donors who stepped in to support additional Tufts students.

Carissa Wang, E26, was one of three Tufts DIAMONDS students supported by a private anonymous donor this year. Her work focused on hospital early warning systems that alert staff when a patient needs support. Current systems have a high rate of false alarms which can overwhelm already-busy hospital staff and create distrust in the system. Using real data from 31,000 intensive care unit stays, Wang helped create a model that could reduce the fraction of unnecessary alarms while still issuing alerts that could improve patient care. Assistant Professor Mike Hughes mentored Wang, and their work could lead to more reliable early warning systems in medical settings.

Funded by a private anonymous donor, computer engineering student Hunter Niimi, E27, spent the summer streamlining reinforcement learning for machine learning (ML) agents. Reinforcement learning trains machine learning agents through trial and error. Using the PLACE (People Leveraging Augmented Heuristics to Change the Environment) method developed by PhD candidate Brennan Miller-Klugman, Niimi and faculty mentor Associate Professor Jivko Sinapov investigated whether adding visuals to training situations would help people train ML agents more efficiently. Their work could enhance ML agent training, making machine learning more efficient in self-driving cars, food delivery robots, and more.

Gaby Fuki, A26, also focused on reinforcement learning in her project. A private anonymous donor supported her work on the multigrid algorithm with Ada Lovelace Associate Professor Hari Sundar and PhD candidate Budvin Edippuliarachichi. Fuki extended previous 2D work on the multigrid algorithm to explore how reinforcement learning agents can be used in simple 3D cases. She helped retrain one machine learning agent to perform graph coarsening on simple 3D graphs and investigated how the agent might work on larger, more complex graphs.

Ultimately the DIAMONDS program showcases Tufts’ strength in data science research. The program facilitators are grateful that anonymous donors and leftover NSF funds came together to enable a breadth of data science projects this summer. DIAMONDS students made valuable connections with faculty members and PhD students as they contributed to cutting-edge research at Tufts. Looking forward, the program plans to continue providing transformative research opportunities for undergraduate students in data science. 

Learn more about the DIAMONDS Program.