About

I am Soumadeep /ˌʃoʊ.moʊˈðiːp/ (pronounced Show-mo-deep), and I completed my PhD in Computer Science at the Computer Vision and Pattern Recognition Unit, ISI Kolkata, under the supervision of Dr. Utpal Garain.

My doctoral research focused on integrating key symbolic AI principles—such as domain constraint awareness and logical coherence—into deep learning systems. Currently, I am exploring techniques to better understand and enhance reasoning in Large Language Models (LLMs), with a focus on constraint adherence, alignment, and reliability.

Recent Works

• PhD Thesis: Domain Obedient Deep Learning

  Soumadeep Saha
  Indian Statistical Institute, Kolkata | 2025

• KisMATH: Do LLMs Have Knowledge of Implicit Structures in Mathematical Reasoning?

  Soumadeep Saha, A. Chaturvedi, S. Saha, U. Garain, N. Asher
  Accepted to TACL (In press) | 2025