Mohammad Junayed Hasan

Graduate Researcher

Johns Hopkins University

Baltimore, MD

I design efficient machine learning and language technologies for healthcare and low-resource settings.

I am a Master’s student in Computer Science at Johns Hopkins University and an AI researcher focusing on efficiency-centric machine learning, natural language processing, and healthcare applications. I am interested in models that reason, translate, and make predictions under realistic compute, memory, and data constraints.

At Johns Hopkins, I am advised by Prof. Anjalie Field and work with Prof. Philipp Koehn on low-resource machine translation and reference-free evaluation, developing reinforcement learning methods that use paraphrase consensus instead of parallel corpora. As a research intern at Mayo Clinic under Dr. Thomas Tavolara, I build end-to-end clinical ML pipelines for blood utilization forecasting and biomarker detection, with an emphasis on deployable systems that integrate into real hospital workflows.

Previously, at North South University, I completed my undergraduate thesis on hybrid quantum-classical machine learning under Prof. Mahdy Rahman Chowdhury, who received the ICO Galileo Galilei Medal Award in 2023. I also worked as an Research Assistant with Prof. Nabeel Mohammed on clinical NLP, smile recognition, and deep metric learning projects including OptimCLM and Shadow Loss. I currently work with Prof. Mahdy as an AI Research Assistant at Mahdy Research Academy, where I lead projects on model compression, hybrid quantum-classical architectures, and multimodal medical imaging. My work has appeared in venues such as NeurIPS, IEEE ICDM, Pattern Recognition Letters, International Journal of Medical Informatics, PLOS ONE, and IEEE Access, along with several manuscripts currently under review at top-tier conferences and journals.

Current Positions

Graduate Researcher, Center for Language and Speech Processing (CLSP), Johns Hopkins University
Data Science / AI Research Intern, Mayo Clinic (Transfusion Medicine & Pathology)
AI Research Assistant & Mentor, Mahdy Research Academy (remote)

Research Areas

Efficient ML & Model Compression
Knowledge distillation, pruning, quantization for edge deployment

Low-Resource & Multilingual NLP
Machine translation, reference-free evaluation, cross-lingual learning

Clinical Decision Support & Medical Imaging
Healthcare AI, biomarker detection, clinical outcome prediction

Hybrid Quantum-Classical Learning
Knowledge transfer to quantum neural networks, efficient quantum architectures

news

Dec 23, 2025	🎓 Graduated from Johns Hopkins University with an M.S.E. in Computer Science (CGPA of 3.9/4.0). Grateful for an incredible journey in research and learning.
Dec 03, 2025	Reached 50+ citations on Google Scholar! 🎉 Grateful for the research community’s engagement with my work.
Dec 03, 2025	QuantumMedKD paper published online in Alexandria Engineering Journal (Vol. 134C, pp. 49-68)! 🎉 A hybrid quantum-classical knowledge distillation framework for medical image analysis. DOI
Dec 01, 2025	Paper on NCD detection via prompt engineering and domain knowledge integration published in Alexandria Engineering Journal! 📄
Nov 15, 2025	Presented HadaSmileNet at IEEE ICDM 2025 (Oral + Poster)! 🎤 “Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles.”
Sep 22, 2025	Paper accepted at Women in Machine Learning Workshop @ NeurIPS 2025! 🎉 “Leveraging ML and LLMs for Enhanced Occupational Stress Detection.”
Sep 22, 2025	CQ-CNN paper accepted in PLOS ONE! 🧠 A lightweight hybrid classical-quantum CNN for Alzheimer’s disease detection using 3D structural brain MRI.

selected publications

AEJ

QuantumMedKD: A hybrid quantum-classical knowledge distillation framework for medical image analysis

MD Nahid Hassan Nishan^{^†}, Mohammad Junayed Hasan^{^†}, and M.R.C. Mahdy

Alexandria Engineering Journal, Jan 2026

Abs DOI Bib HTML

@article{nishan2026quantummedkd,
  title = {QuantumMedKD: A hybrid quantum-classical knowledge distillation framework for medical image analysis},
  author = {Nishan<sup></sup>, MD Nahid Hassan and Hasan<sup></sup>, Mohammad Junayed and Mahdy, M.R.C.},
  journal = {Alexandria Engineering Journal},
  volume = {134C},
  pages = {49--68},
  year = {2026},
  month = jan,
  publisher = {Elsevier},
  doi = {10.1016/j.aej.2025.12.007},
  dimensions = {true}
}

ICDM 2025
HadaSmileNet: Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles

Mohammad Junayed Hasan, Nabeel Mohammed, Shafin Rahman, and 1 more author

IEEE International Conference on Data Mining (ICDM), Sep 2025

ICDM 2025 Abs arXiv Bib HTML Video Code Poster

Accepted for Oral + Poster Presentation

This paper presents HadaSmileNet, a novel approach for spontaneous smile recognition using Hadamard fusion of handcrafted and deep learning features. Accepted for oral and poster presentation at IEEE ICDM 2025.
@article{hasan2025hadasmilenet, title = {HadaSmileNet: Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles}, author = {Hasan, Mohammad Junayed and Mohammed, Nabeel and Rahman, Shafin and Koehn, Philipp}, journal = {IEEE International Conference on Data Mining (ICDM)}, year = {2025}, month = sep, publisher = {IEEE}, }
IJMI
OptimCLM: Optimizing clinical language models for predicting patient outcomes via knowledge distillation, pruning and quantization

Mohammad Junayed Hasan, Fuad Rahman, and Nabeel Mohammed

International Journal of Medical Informatics, Jan 2025

Abs DOI Bib HTML Code

We present OptimCLM, a comprehensive framework for optimizing clinical language models through knowledge distillation, pruning, and quantization. Our approach reduces model size by 95% while maintaining 98% of the original performance on clinical NLP tasks, achieving 96.5% reduction in inference latency.
@article{hasan2025optimclm, title = {OptimCLM: Optimizing clinical language models for predicting patient outcomes via knowledge distillation, pruning and quantization}, author = {Hasan, Mohammad Junayed and Rahman, Fuad and Mohammed, Nabeel}, journal = {International Journal of Medical Informatics}, volume = {195}, pages = {105764}, year = {2025}, month = jan, publisher = {Elsevier}, doi = {10.1016/j.ijmedinf.2024.105764}, dimensions = {true}, }

PRL

DeepMarkerNet: Leveraging supervision from the Duchenne Marker for spontaneous smile recognition

Mohammad Junayed Hasan, Kazi Rafat, Fuad Rahman, and 2 more authors

Pattern Recognition Letters, Dec 2024

Abs DOI Bib HTML Code

@article{hasan2024deepmarkernet,
  title = {DeepMarkerNet: Leveraging supervision from the Duchenne Marker for spontaneous smile recognition},
  author = {Hasan, Mohammad Junayed and Rafat, Kazi and Rahman, Fuad and Mohammed, Nabeel and Rahman, Shafin},
  journal = {Pattern Recognition Letters},
  volume = {186},
  pages = {148--155},
  year = {2024},
  month = dec,
  publisher = {Elsevier},
  doi = {10.1016/j.patrec.2024.09.015},
  dimensions = {true},
}

Under Review
Shadow loss: Memory-linear deep metric learning for efficient training

Alif Elham Khan^{^†}, Mohammad Junayed Hasan^{^†}, Humayra Anjum^{^†}, and 1 more author

CVPR 2026 - Under Review, Dec 2025

Abs arXiv Bib HTML

We introduce Shadow Loss, reducing memory complexity from O(N²) to O(N) for deep metric learning while accelerating convergence 1.5-2×. Validated on CUB-200, CARS-196, and large-scale retrieval datasets.
@article{khan2025shadow, title = {Shadow loss: Memory-linear deep metric learning for efficient training}, author = {Khan, Alif Elham and Hasan, Mohammad Junayed and Anjum, Humayra and Mohammed, Nabeel}, journal = {CVPR 2026 - Under Review}, year = {2025}, }