Building trustworthy multimodal AI

Quick highlights

Peer-reviewed journals: 7 · Under review/submitted: 12 · Selected conference papers: 10 · Best Paper awards: 5

• Research Assistant at East West University (AI Lab) and Ulster University (School of Computing).
• Work spans multimodal learning, explainability, and deployment feasibility (edge/mobile).
• Conference presentations and invited talks across multiple venues.

• Developing a semi-supervised self-attention model for calibrated multimodal fusion to enable real-time glaucoma detection in clinical settings.

• Building a retrofit computer-vision system for real-time fabric fault detection and quality grading, with a web-based dashboard for factory deployment.

• Developed a psychophysics-driven EEG framework for early detection and risk stratification of anxiety and depression among EWU students.

• Proposed XSVC (XceptionNet+SVC) for 10-class tomato leaf disease identification using 60,000 augmented images, achieving 99.51% accuracy, and deployed an Android predictor app.

Research Assistant, East West University Dhaka, Bangladesh

Artificial Intelligence Lab, Supervisor: Prof. Dr. Ahmed Wasif Reza Jan 2022 – Present

• Conducted and presented research at over ten international conferences (IEEE and Springer) across India, Italy, Nepal, and Bangladesh, including multiple Best Paper Awards.
• Supervised and trained more than 200 hundred undergraduate and graduate researchers in AI experimentation, data analysis, and scientific publishing, fostering collaboration across disciplines.
• Designed and deployed Vision Transformer–based systems for medical, agricultural, and educational applications, improving automation and diagnostic precision.
• Integrated AI-driven control and IoT systems to enhance smart infrastructure efficiency and enable real-time autonomous operation.

Research Assistant, Ulster University Coleraine, Northern Ireland

School of Computing, Supervisor: Dr. Nazmul Siddique Jan 2023 – Present

• Investigating multimodal data fusion methods for improved object recognition using large-scale datasets and hybrid Transformer–CNN architectures.
• Developed data preprocessing pipelines that reduced multimodal noise and improved data consistency.
• Implemented optimized feature fusion models capturing both low-level interactions and high-level semantic relationships across visual, audio, and textual modalities.
• Contributed to the creation and annotation of large multimodal datasets for benchmarking feature extraction and model evaluation.

Research Assistant, East West University Dhaka, Bangladesh

Center for Research and Training (CRT), Supervisor: Dr. Fakir Mashuque Alamgir Oct 2025 – Present

• Conducting research on multimodal fusion and transformer-based deep learning models for early detection of glaucoma.
• Developing and evaluating an explainable AI (XAI) framework to improve interpretability and clinical trust in medical image diagnostics.
• Designing a mobile application integrating the trained model for real-time glaucoma risk prediction and screening.