Welcome!

I am a PhD student in Computer Science at the University of Tennessee, Knoxville under the guidance of Prof. Dongsheng Ding. Previously, I developed deep learning models for 3D medical images at ImmersiveTouch. During my graduate studies at Rensselaer Polytechnic Institute, I worked on Natural Language Processing and Machine Learning. Before that, I graduated from Bangladesh University of Engineering and Technology with a B.Sc. in Electrical and Electronic Engineering.

Current Research

I work on making generative models more controllable and predictable My research lies at the intersection of generative models, constrained optimization, and representation learning. Specifically, I study constrained optimization methods for controlling diffusion models, enforcing desired behaviors and suppressing unwanted ones, so that guidance weights are derived in a principled, interpretable way rather than tuned by hand.

Past Experience

ImmersiveTouch: 3D Medical Imaging
I worked on segmentation and landmark detection models for CT scans, developing methods to automatically delineate anatomical structures and generate 3D reconstructions of organs to support surgical planning.

Rensselaer Polytechnic Institute: NLP & Representation Learning
I explored problems in representation learning, model efficiency, and information extraction from text:

• Domain adaptation for causal information extraction - studying how neural models trained on one domain could generalize to extract causal relations in another.
• Multi-sense word embeddings - addressing the meaning conflation problem in traditional word embeddings, including a knowledge distillation approach to transfer contextual information from large pre-trained language models into multi-sense embedding spaces.

IBM Research Internships
I worked on information extraction from structured documents, including scanned images and PDFs, contributing to systems for automated document understanding.