Alum Alexander Levine Honored with Charles A. Caramello Distinguished Dissertation Award

University of Maryland Department of Computer Science alum Alexander Levine (Ph.D. '23, computer science) has been awarded the Charles A. Caramello Distinguished Dissertation Award for his dissertation titled "Scalable Methods for Robust Machine Learning." Levine, now a postdoctoral fellow at the University of Texas at Austin, focused on developing machine learning models that maintain accuracy amid distortions. The award ceremony is scheduled for Tuesday, May 14, at the Stamp Student Union. The award is for the dissertation he completed in 2023.

The Charles A. Caramello Distinguished Dissertation Award is given annually by the Graduate School to recognize dissertations that provide highly original contributions that make an unusually significant contribution to the discipline. Levine is among four recipients of the award this year.

Awardees receive an honorarium of $1,000. Additionally, they may be nominated for further recognition at the national level through the CGS/ProQuest Distinguished Dissertation Award competition, which selects outstanding dissertations from across the country to honor achievements in graduate research.

“I feel honored that my work has been recognized by this award,” Levine said. “I am deeply thankful for all of the support I received during my time at UMD from my advisor, my collaborators, my dissertation committee and the rest of the UMD computer science community. I am fortunate to have worked with such talented people on such interesting problems.”

Advised by Associate Professor Soheil Feizi, Levine's dissertation introduces innovative methods for ensuring the robustness of machine learning models, specifically in scenarios where input data may be subtly altered or distorted, including malicious tampering. This research is particularly relevant as machine learning applications become increasingly prevalent in areas requiring high reliability and security.

Levine explained that practitioners can implement these systems more confidently in safety-critical applications by developing machine learning techniques with well-understood robustness guarantees. He noted that the capabilities of machine-learning-based systems have expanded dramatically in just the last couple of years, increasing their use in various sectors. Levine emphasized the growing importance of ensuring these systems' robustness as their applications broaden.

Levine is currently expanding his research focus.

“At UT Austin, my research focus has shifted to representation learning for sequential decision-making tasks,” Levine shared. “In particular, I have been working on frameworks that allow deep learning to be used in combination with search-based planning techniques, so that we can benefit from both the powerful capabilities of modern deep learning and the interpretability, flexibility and efficiency of classical planning methods.”

Levine received the Larry S. Davis Doctoral Dissertation Award in the Fall of 2023. Named in honor of Computer Science Professor Emeritus Larry Davis, the award, given by UMD’s Department of Computer Science, highlighted dissertations that were exceptional in their technical depth and potential for significant impact.

—Story by Samuel Malede Zewdu, CS Communications