My research lies at the intersection of Natural Language Processing (NLP), Deep Learning, and Human-Centered AI, with a focus on low-resource languages such as Bengali. As a founding member and active researcher at KCR-Lab, I supervise student research projects and lead several initiatives addressing real-world problems in e-commerce, healthcare, and digital discourse analysis.

I am particularly passionate about building benchmark datasets, exploring multilingual transformer architectures, and deploying AI solutions that work effectively in underrepresented regions. My work integrates modern deep learning techniques like BERT, Vision Transformers, and hybrid models in NLP, CV, and beyond.

Current research themes include:

• Natural Language Processing for Low-Resource Languages (Bengali sentiment, emoji-aware reviews)
• AI for Social Good (filtering sports comments containing hate/gambling content)
• Deep Learning in E-commerce Feedback Analytics
• Vision Transformers for Healthcare (fundus image classification)
• Dataset development and annotation for applied machine learning

Research Experience

Founding Member & Researcher, KCR-Lab
📍 Oct 2024 – Present

• Supervise undergraduate research in machine learning, deep learning, and computational linguistics
• Organize workshops and hands-on training sessions to equip students with AI skills
• Mentor students in research methodology, experimental design, and academic writing
• Lead NLP projects focusing on low-resource languages, especially Bengali
• Collaborate with faculty on interdisciplinary AI initiatives

Undergraduate Thesis Researcher, CUET
📍 Feb 2022 – Apr 2023
Title: A Deep Learning Based Pharmaceutical Product Evaluation from Bengali Reviews Considering Emoji

• Developed a benchmark dataset of 4,000+ Bengali pharmaceutical reviews
• Identified the rating–review mismatch problem in local e-commerce
• Proposed a novel emoji-aware approach to enhance review classification
• Applied traditional ML and hybrid deep learning models for sentiment analysis
• Supervised by Prof. Dr. Muhammad Ibrahim Khan

[C.1] BScFilter: A Deep Learning Approach for Sports Comments Filtering in a Resource-Constrained Language
A. Rahman, M.I. Khan, M.M.H. Rifat
5^th International Conference on Trends in Computational and Cognitive Engineering (TCCE 2023, Springer) [Paper]

[C.2] Vision Transformers for Multi-Class Eye Disease Classification: Enhancing Early Detection in Resource-Constrained Healthcare
A. Rahman, M. Hasan, M.A. Mia, M.I. Khan
3^rd International Conference on Big Data, IoT and Machine Learning (BIM 2025, Springer) [Paper]

[C.3] Addressing the Mental Health Crisis: Understanding Suicidal Risk Factors in University Students Through Interpretable Machine Learning
A. Islam, A. Rahman, M.J. Abdin, O. Rahaman, M.N. Alam
International Conference on Intelligent Data Analysis and Applications (IDAA 2025, Springer)

[C.4] SETBoost: An Interpretable Machine Learning Approach for Predicting Software Employee Turnover Tendency
A. Rahman, M.I. Hossain, M.A. Mia, M.I. Khan
28^th International Conference on Computer and Information Technology (ICCIT 2025, IEEE Xplore) [Preprint]

[C.5] Evaluating Prompting and Fine-Tuning Approaches for Bengali Violence Detection in a Low-Resource Language
M.A. Mia, A. Hossain, A. Rahman, M.S. Mursalin, M.I. Khan
28^th International Conference on Computer and Information Technology (ICCIT 2025, IEEE Xplore) [Preprint]

Works in Progress

[J.1] BanGRev: A novel Bengali tech gadget review corpus utilizing triple-pooling feature fusion with explainable AI
M.A. Mia, A. Rahman, M.I. Khan, I.H. Sarker
📝 Journal Submission – Under Review

[J.2] Addressing Rating-Review Discrepancy in a Novel Dataset: An Explainable Pharmaceutical Product Evaluation using Multi-Stream Attention Transformer
A. Rahman, M.A. Mia, M.I. Khan, I.H. Sarker
📝 Journal Submission – Under Review

Research Projects

Diabetes Factor Analysis with Outlier Detection [GitHub]
Technologies: Scikit-learn, pandas, numpy, matplotlib, seaborn

• Applied statistical methods to identify significant factors affecting diabetes diagnosis
• Implemented outlier detection techniques to improve feature selection quality
• Employed ensemble methods (Random Forest, XGBoost) for enhanced prediction accuracy
• Achieved 2–4% overall performance increment with outlier removal

Sentiment Analysis on Twitter Texts [GitHub]
Technologies: Scikit-learn, NLTK, Transformers, Tkinter, pandas, matplotlib

• Developed a system for social media sentiment classification
• Implemented multiple feature extraction techniques (TF-IDF, word embeddings)
• Compared performance across classical ML and deep learning approaches
• Created interactive interface for real-time sentiment prediction

CUET-BUS-TRACKER: Real-time Transportation Monitoring System [GitHub]
Technologies: Java, Firebase, Google Maps API, Android Studio

• Designed and implemented location tracking system for university transport
• Built Android application with real-time updates for student convenience
• Integrated Google Maps API with Firebase for synchronous data updates

📌 For further details, check out my Google Scholar profile.