I want to build diagnostic AI systems that work reliably when labeled medical data is scarce — combining semi-supervised learning, explainability, and computational efficiency to make deep learning trustworthy enough for real clinical use.

I am a CSE undergraduate at RUET (Rajshahi University of Engineering & Technology), Bangladesh. My current work centers on semi-supervised learning for retinal OCT image classification, where I develop attention-based models that reduce dependence on large annotated datasets. This sits alongside two other threads from my published work: an explainable multi-scale CNN framework for Alzheimer's diagnosis from MRI, and a memory-efficient segmentation study for resource-constrained deployment. Together these reflect the direction I want a PhD to sharpen: models that are not just accurate, but data-efficient, interpretable, and deployable in real healthcare settings.