About Me
I am an Assistant Professor of Computer Science and Engineering at NPI University of Bangladesh, where I founded and lead the CARMA AI Lab. My research builds trustworthy AI for clinical decision-making.
My work is driven by a simple observation from the clinic: a model that is accurate on a benchmark can still fail a patient when the hospital, the scanner, or the population changes. I build methods that catch those failures before they reach the bedside.
My Path
I received my B.Sc. in Computer Science and Engineering from the Bangladesh University of Business and Technology (BUBT) in 2021, and stayed on as a lecturer. There I mentored undergraduate research projects in medical imaging and computer vision — several of those student projects grew into peer-reviewed publications — and served as advisor to the IEEE CS BUBT student branch chapter.
I then earned my M.S. in Computer Science from Kent State University (2025), where I worked as a graduate teaching assistant while researching trustworthy machine learning for medical image analysis.
In 2026 I returned to Bangladesh to join the faculty of NPI University of Bangladesh as an Assistant Professor, where I also head the ICT Cell and founded the CARMA AI Lab.
Research
My research starts with medical imaging. I build deep learning systems for chest X-ray, MRI, and CT analysis — designed not for leaderboards, but for the clinic, where data shifts across hospitals, scanners, and patient populations. Everything else I work on serves that goal: making sure these models are reliable when it matters.
That reliability question breaks into four threads I pursue in parallel:
- Evaluation & Auditing. Forensic auditing of ensemble classifiers: calibration, shortcut detection, and failure analysis when models meet real clinical data.
- Generalization. Predicting out-of-distribution performance without target data. How do medical vision models behave across hospitals, scanners, and populations?
- Interpretability. Human-centered explanations for clinical decision support: what the model looked at, and why a clinician should (or should not) trust it.
- Safety & Uncertainty. Post-hoc uncertainty quantification for frozen black-box vision models. Knowing when to predict, when to abstain, and when to refer.
Two active threads drive this work right now: STAMP, which predicts out-of-distribution performance without target data, and SpatialUQ, which quantifies uncertainty for frozen black-box vision models from spatial consistency. Both are currently under review.
Selected Publications
Under ReviewSubmitted to ICLR 2027
STAMP: Predicting Out-of-Distribution Generalization without Target Data
M. K. Mahbub and M. Biswas
Estimates how a trained model will perform at a new hospital using only source-domain images — no target data or labels needed.
BibTeX
@misc{mahbub2026stamp,
title={STAMP: Predicting Out-of-Distribution Generalization without Target Data},
author={Mahbub, Md Kawsher and Biswas, Milon},
year={2026},
eprint={2609.32672},
archivePrefix={arXiv},
primaryClass={cs.LG}
}Under ReviewSubmitted to AISTATS 2027
SpatialUQ: Post-Hoc Uncertainty Quantification from Spatial Consistency in Black-Box Vision Models
M. K. Mahbub, M. Biswas, M. N. Morshed, and W. Yu
Flags unreliable predictions of frozen black-box vision models from spatial consistency — six forward passes, no retraining, no internals.
BibTeX
@misc{mahbub2026spatialuq,
title={SpatialUQ: Post-Hoc Uncertainty Quantification from Spatial Consistency in Black-Box Vision Models},
author={Mahbub, Md Kawsher and Biswas, Milon and Morshed, Mirza Niaz and Yu, Wei},
year={2026},
eprint={2610.09498},
archivePrefix={arXiv},
primaryClass={cs.CV}
}JournalInformation Sciences
Deep features to detect pulmonary abnormalities in chest X-rays due to infectious diseases: COVID-19, pneumonia, and tuberculosis
Mahbub, M. K., Biswas, M., Gaur, L., Alenezi, F., & Santosh, K. C.
Deep-feature pipeline detecting COVID-19, pneumonia, and tuberculosis from chest X-rays — my most-cited paper.
BibTeX
@article{mahbub2022deep,
title={Deep features to detect pulmonary abnormalities in chest X-rays due to infectious diseases: COVID-19, pneumonia, and tuberculosis},
author={Mahbub, Md Kawsher and Biswas, Milon and Gaur, Loveleen and Alenezi, Fayadh and Santosh, K. C.},
journal={Information Sciences},
volume={592},
pages={389--401},
year={2022},
publisher={Elsevier},
doi={10.1016/j.ins.2022.01.062}
}Full publication list →
Recent News
Oct 2026milestoneFounded the CARMA AI Lab — Clinical AI for Reliable Medical Analytics.
Oct 2026paperSpatialUQ: post-hoc uncertainty quantification from spatial consistency — submitted to AISTATS 2027.
Sep 2026paperSTAMP: predicting out-of-distribution generalization without target data — submitted to ICLR 2027.
2026paper'Foundation Models in Healthcare' survey under review at ACM Computing Surveys.
All news →
Appointments
Oct 2026 – Present
Founder & Director
CARMA AI Lab — Clinical AI for Reliable Medical Analytics
Leading research on reliable clinical AI across four pillars: evaluation and auditing, OOD generalization, interpretability, and safety and uncertainty.
Mar 2026 – Present
Assistant Professor
NPI University of Bangladesh, Dept. of CSE
Teaching algorithms and AI. Head of the ICT Cell since April 2026.
Aug 2023 – Dec 2025
Graduate Teaching Assistant
Kent State University, OH, USA
Assisted graduate CS courses; conducted research in medical image analysis and trustworthy machine learning.
Dec 2021 – Jul 2023
Lecturer
Bangladesh University of Business and Technology (BUBT), Dhaka
Taught undergraduate CS courses; Advisor, IEEE CS BUBT Student Branch Chapter.
Education
Aug 2023 – May 2025
M.S. in Computer Science
Kent State University, OH, USA
CGPA 3.80 / 4.00
Feb 2017 – May 2021
B.Sc. in Computer Science and Engineering
Bangladesh University of Business and Technology (BUBT)
CGPA 3.84 / 4.00
Academic service
Program Committee Member
IEEE CBMS International Symposium on Computer-Based Medical Systems
2023 – 2026
Program Committee Member
ICANN 2026 — 35th International Conference on Artificial Neural Networks
2026
Reviewer
Expert Systems (Wiley)
Ongoing
Reviewer
CMC – Computers, Materials & Continua (Tech Science Press)
Ongoing
Reviewer
International Journal of Intelligent Transportation Systems Research (Springer)
Ongoing
Awards & Honors
First Place and People's Choice Award, International Cook-Off 2024
Kent State University
2024
Merit-based Scholarship (5 times)
Bangladesh University of Business and Technology (BUBT)
Download full CV (PDF)
Prospective collaborators: I am always happy to discuss research collaborations — especially on auditing, generalization, and uncertainty in medical AI. Email me with a short note about the problem you want to work on.