PhD Spotlight: Aria Masoomi, PhD’24, Computer Engineering, and Mathematics
Aria Masoomi, PhD’24, computer engineering, and PhD’24, mathematics, focused his dissertation research for computer engineering on “Making Deep Neural Networks Transparent.” His mathematics thesis was “Poisson Geometry of Flag Varieties and Representation Theory of Their Quantum Deformation.”
Aria Masoomi, PhD’24, began the PhD program in computer engineering at Northeastern in 2017, while simultaneously pursuing a PhD in mathematics. His computer engineering research thesis was titled “Making Deep Neural Networks Transparent.” His mathematics thesis was “Poisson Geometry of Flag Varieties and Representation Theory of Their Quantum Deformation” for which he earned the Best Thesis Award. He was advised by Jennifer Dy, professor of electrical and computer engineering, and Milen Yakimov, professor of mathematics.
Masoomi’s research lies at the intersection of theoretical machine learning and modern algebra. On the machine learning (ML) side, he develops mathematically grounded algorithms that uncover feature interactions and information flow in deep networks. This work has led to four NeurIPS papers, including a Spotlight selection, and additional publications at ICLR and AISTATS, covering topics ranging from explainable artificial intelligence methods to foundational deep learning theory. He applies these advances in biomedicine, focusing on chronic obstructive pulmonary disease (COPD). His ML methods address COPD heterogeneity and have appeared in multiple journal articles. For mathematics, Masoomi studies the representation theory of quantum function algebras on flag varieties and the Poisson geometry of their semiclassical limits. In his recent work, they resolve the Poisson degeneracy locus for a broad class of flag varieties, strengthening connections between non-commutative algebraic geometry and Poisson geometry. Masoomi is an active reviewer for NeurIPS, AISTATS, and AAAI and has delivered invited talks at Harvard’s Data to Actionable Knowledge Lab and the American Mathematical Society.
In support of the Northeastern community, Masoomi co-organized the Machine Learning Summer School for Mathematicians and Physicists in Northeastern’s math department, helping PhD students integrate ML into their research. By uniting rigorous theory with real-world data, Masoomi aims to build machine-learning systems that are interpretable, robust, and mathematically elegant.