CV

Download Full CV (PDF)

Education

Khulna University of Engineering and Technology (KUET) — Khulna, Bangladesh B.Sc. in Electrical and Electronic Engineering (Jan 2019 – Mar 2024) CGPA: 3.71 / 4.00 — ranked 24 of 149 (top 16%) Thesis: Assessing Mental Stress Using Deep Learning Methods from EEG Signals Advisor: Dr. Mohiuddin Ahmad, Professor of EEE

Research Interests

• One- and few-step generative modeling: closed-form flow and transport formulations that collapse iterative sampling into a single network evaluation
• Image restoration and exposure correction under real-world degradation, formulated through principled optimal transport rather than heuristic reconstruction losses
• Geometry-aware deep learning on Riemannian and SPD manifolds: equivariant architectures, geodesic operators, and numerically stable matrix-function approximation
• Structured representation learning for neural and biomedical signals, including EEG-based affective and cognitive state decoding

Research Experience

Researcher

November 2024 – Present

• Derived a closed-form flow formulation that collapses iterative diffusion sampling into a single network evaluation, removing the multi-step solver required by conventional flow matching; implemented the complete reference training codebase with stabilizers for the degenerate-path regime.
• Proposed Equivariant Geodesic Networks (EGN), an end-to-end classifier of equivariant mapping, Riemannian pooling, geometric bias, and geodesic attention layers operating directly on the SPD manifold.
• Derived orthogonal-polynomial approximations for matrix-logarithm normalization in global covariance pooling, reducing cost and enabling GPU execution of a previously CPU-bound operation.
• Designed AutoLumNet, a bi-branch exposure-aware network for single-shot correction of under- and over-exposed images; reformulated correction as monotone optimal transport (average MSEC 23.75).
• Developed DAFNet, integrating a local 2D-CNN branch with channel attention and a global graph-attention transformer with covariance pooling, achieving state-of-the-art EEG emotion recognition.

Research Collaborator — EEE, KUET

Supervisor: Dr. Md. Mahbub Hasan · May 2024 – July 2024

• Constructed the first EEG auditory-evoked-potential dataset for imagined Bengali language tasks using a low-cost OpenBCI system.
• Trained and evaluated convolutional models for Bengali vowel and word recognition; the resulting paper received the IEEE Best Paper Award.

Undergraduate Thesis Researcher — EEE, KUET

Advisor: Dr. Mohiuddin Ahmad · January 2023 – March 2024

• Established how time–frequency features and kernel scale interact in 3D CNNs for EEG emotion recognition.
• Built the complete preprocessing, class-balancing, and evaluation pipeline on constrained hardware without GPU access.

Publications

Refereed Journal Articles

1. DAFNet: A Dual-Path Attention Fusion Network for EEG Emotion Recognition via CNN and Graph-Based Global Modeling. MR Khan, AA Tania, TA Bushra, J Pritha, M Ahmad. Array (Elsevier), 2025.
2. A Comparative Study of Time–Frequency Features Based Spatio-Temporal Analysis with Varying Multiscale Kernels for Emotion Recognition from EEG. MR Khan, AA Tania, M Ahmad. Biomedical Signal Processing and Control (Elsevier), 2025.
3. Multiclass Liver Disease Prediction with Adaptive Data Preprocessing and Ensemble Modeling. AA Ahad, B Das, MR Khan, N Saha, A Zahid, M Ahmad. Results in Engineering (Elsevier), 2024.

Refereed Conference Proceedings

1. Orthogonal Polynomial Approximation for Matrix-Log Normalization in Global Covariance Pooling. MRU Rahman, MR Khan, MSH Shovon, P Liò, MA Moni. BMVC 2026 (accepted).
2. Recognition of Bengali Vowels from Auditory Evoked Potentials Using CNN. T Das, MR Khan, MM Hasan. IEEE SPICSCON 2024 — Best Paper Award.
3. Custom Dataset-Driven Unsupervised Low-Light Image Enhancement Using 2D CNN. AA Tania, MR Khan, M Ahmad. IEEE QPAIN 2025.
4. Mental Stress Detection from EEG Signals Using Comparative Analysis of Random Forest and Recurrent Neural Network. MR Khan, M Ahmad. IEEE iCACCESS 2024.
5. Machine Learning Techniques for Brain Stroke Analysis and Prediction. R Hasan, SMR Islam, MR Khan. IEEE SPICSCON 2024.
6. Smart Classroom Automation: A Fusion of AI with Voice, Gesture, and Face Recognition Attendance System. MR Khan, AA Ahad, AA Tania, T Das, B Das. IEEE iCACCESS 2024.

Manuscripts Under Submission

1. AutoLumNet: Monotone Optimal Transport for Single-Shot Exposure Correction. AA Tania, MR Khan, M Ahmad. IEEE TPAMI (under review, 2026).
2. Equivariant Geodesic Networks: End-to-End Classification on the SPD Manifold. MR Khan, AA Tania. AAAI 2026 (under review).
3. MCFN: A Lightweight Multi-Path Compound Fusion Network for Deepfake Face Detection. SS Lagna, MR Khan. Neural Computing and Applications, Springer (under review, 2026).
4. Trigonometric B-Spline Convolutional KAN for Texture Classification. MRU Rahman, MR Khan, R Islam, M Akter. ACCV 2026 (under review).
5. Difficulty-Calibrated Interpolation Paths for Conditional Flow Matching. AA Tania, MR Khan. IEEE ICCAS 2026 (under review).

Software & Open Source

egnlib — Equivariant Geodesic Networks for the SPD manifold (2026) Open-source Python package implementing the full EGN architecture (equivariant mapping, Riemannian pooling, geometric bias, geodesic attention) with a stable matrix-logarithm backward pass. pip install egnlib · source: github.com/kraihan/EGN

Teaching Experience

Lecturer, Department of EEE — North Western University, Khulna, Bangladesh (August 2024 – Present)

• Digital Image Processing (EEE-4255) — lectures, labs, assignments, and computer-vision project supervision.
• Digital Signal Processing (EEE-4115) — theory and biomedical signal analysis, including the DFT and FFT.
• Sessional on Digital Signal Processing (EEE-4116) — laboratory sessions and EEG signal-processing projects.
• Basic Electrical Circuits (EEE-1221) — fundamental laws, network theorems, and circuit analysis.
• Supervise three undergraduate thesis students from problem formulation through manuscript submission (programming, classical ML, deep learning, signal processing).

Academic Service

• Batch Advisor, third-year cohort, Department of EEE, North Western University (2025 – Present)
• Course Coordinator, Department of EEE, North Western University (2024 – Present)
• Co-founder, departmental research group, North Western University (2024 – Present)
• Vice President, KUET Mathematics Club (Feb 2023 – Feb 2024)
• Ambassador, International Youth Math Challenge — mathematics outreach in Bangladesh (2021)

Honors & Awards

• Honorable Mention, International Mathematics Competition (IMC), Bulgaria — 589 participants worldwide (2021)
• First participant from Bangladesh, 27th International Mathematics Competition (2020)
• Gold Medal, International Youth Math Challenge — top 1% of 15,000 competitors (2020, 2021)
• Excellence Award, IYMC Ambassador Programme — fifth among 580 ambassadors (2021)
• Best Paper Award, IEEE SPICSCON (2024)
• Silver & Bronze Medals, International Astronomy and Astrophysics Competition (2021, 2022)
• Best Project Award, Electrical and Electronic Project Design Showcase, KUET (2023)
• University Dean’s Scholarship, KUET — second and fourth academic years (2021, 2024)
• University Merit Scholarship, KUET — awarded seven times for departmental ranking (2020–2024)

Technical Skills

• Programming: Python, C++, MATLAB
• Frameworks & Libraries: PyTorch, TensorFlow, Keras, scikit-learn, NumPy, pandas, SciPy, OpenCV, MNE, EEGLAB, Matplotlib, Seaborn, Flask
• Tools: Git & GitHub, LaTeX, Linux
• Methods: Riemannian / SPD-manifold learning, optimal transport, flow matching & diffusion models, graph neural networks, EEG signal processing, low-level computer vision
• Languages: Bengali (native), English (professional working proficiency)

References

Dr. Mohiuddin Ahmad — Professor, EEE, KUET Email: ahmad@eee.kuet.ac.bd · (undergraduate thesis advisor)

Dr. Kalyan Kumar Halder — Professor, EEE, KUET Email: kalyan@eee.kuet.ac.bd · (course instructor and IMC mentor)