Md Fokhrul Islam

PhD Student & RA · School of Data Science, UVA

I am a Data Science PhD student at the University of Virginia, advised by Professor Heman Shakeri. I work on machine learning for complex, safety-critical systems — healthcare technologies, robots acting in the physical world, and the societal systems they touch.

The question that drives my research: how do we build AI systems that are genuinely intelligent, rather than just well-trained? Most of today's models get their performance from enormous labeled datasets, but the systems I care about rarely have that luxury — and there, the cost of a mistake is not a benchmark point but more than that. I want models that adapt quickly, generalize to situations they have never seen, and come with guarantees: no matter how unfamiliar the situation.

I am working toward this along three threads:

  • Learning with few or no labels — exploiting the geometry of high-dimensional representation spaces so that models can recognize and reason about things they were never trained on.
  • Safety by design — combining control theory with learned models, so that closed-loop systems like the artificial pancreas can adapt to each patient while provably avoiding harmful states.
  • Intelligence as adaptation — I find the world-model line of thinking compelling: self-supervised predictive representations and planning as optimization, with safety constraints built into the objective. Less a description of where my research sits than a compass for where it is headed.

Before UVA I completed my MSc and BSc in Robotics & Mechatronics Engineering at the University of Dhaka, where I worked with Professor Sejuti Rahman & Professor Shafin Rahman.

news

Aug 2025Started PhD at University of Virginia.
Apr 2025Paper accepted at NeurocomputingReKon3D: Contrastive GAN for Zero-Shot 3D Recognition.
Apr 2024Paper accepted at Image and Vision ComputingVAE-GAN3D: 3D Zero-Shot via Image Semantics.
Jan 2024Joined ACI Limited as Machine Learning Engineer.
Jul 2023Successfully defended MSc thesis.
May 2023Paper accepted at IEEE AccessData-Augmentation for Bangla-English Code-Mixed SA.
Nov 2022Review paper accepted at IEEE TNSREAI-driven Stroke Rehabilitation.
Oct 2022IHABOT placed 2nd at DU Research-Publication Fair.
Aug 2022Attended OxML Summer School 2022.
Jun 2022Undergrad thesis selected for poster at CVPR WiCV 2022.
Feb 2022First paper accepted at IEEE TNSREGCN for Physical Rehab Assessment.

selected publications

IEEE TNSRE 2022
Graph Convolutional Networks for Assessment of Physical Rehabilitation Exercises
Swakshar Deb*, Md Fokhrul Islam*, Shafin Rahman, Sejuti Rahman
IEEE Access 2023
Data-Augmentation for Bangla-English Code-Mixed Sentiment Analysis
Mohammad Tareq, Md Fokhrul Islam, Swakshar Deb, Sejuti Rahman, Abdullah Al Mahmud
see all publications →

latest posts

see all posts →