Mahmudul Hasan
I am currently pursuing my M.Sc. in Computer Science and Engineering at the University of Dhaka and working as a Consultant in the ICSETEP Sub-project (UGC), University of Dhaka, where I contribute to AI-driven enterprise business intelligence systems. I am also actively involved in research activities at the Cognitive Agents & Interaction Lab (CAIL).
My research interests include Artificial Intelligence, Deep Learning, and Natural Language Processing. I focus on developing practical AI solutions for real-world challenges, particularly in low-resource language technologies, advanced OCR systems, and AI-powered educational technologies.
I aspire to pursue a PhD and build a career in academia as a teacher, researcher, and mentor, contributing to impactful AI research and helping develop the next generation of scientists and engineers.
Publications & Presentations
Accepted
Developed a resource-efficient HTR system for Bengali script using grapheme-based tokenization and decoder-only transformers, achieving ~95% accuracy in character-level recognition.
Preprints
Applied transformer-based DETR model for robust vehicle detection in challenging road conditions, addressing safety concerns in transportation.
Google Scholar: View full publication list
ORCID: 0009-0009-3125-1340
Research & Professional Experience
Contributing to an ADB-funded project on enterprise AI systems and data-driven business intelligence platforms.
- Designing scalable AI pipelines integrating enterprise databases with Retrieval-Augmented Generation (RAG)
- Developing intelligent systems for real-time insight, predictive, and prescriptive analytics
- Supporting deployment and optimization of GPU-based compute infrastructure for large-scale AI workloads
- Collaborating with multidisciplinary teams across national projects under UGC
Conducted research in Artificial Intelligence and Machine Learning with a focus on NLP, OCR, and low-resource language technologies.
- Developed deep learning-based OCR systems for Bengali handwritten text recognition
- Trained and optimized models on multi-GPU servers using PyTorch
- Implemented mixed-precision training, gradient checkpointing, and distributed pipelines
- Co-authored research published at EMNLP 2025
- Mentored undergraduate students in AI research and machine learning deployment
Developing backend services and scalable data pipelines using Django, PostgreSQL, Celery, and Redis. Contributing to the development of an Android-based education platform used by teachers across rural Bangladesh.
Taught advanced mathematics for university admission preparation. Mentored students in analytical problem-solving and mathematical reasoning.
Education
Focus: Advanced topics in Artificial Intelligence and Machine Learning
Research: Deep Learning architectures, Transformer models, AI applications
Thesis: Bangla Abstractive Summarizer with Attention-based Learning Technique
for Long Articles
Relevant Coursework: Advanced Algorithms, Data Science, Artificial
Intelligence, Machine Learning, Database Systems, Computer Networks, Discrete Mathematics,
Linear Algebra
GPA: 5.00/5.00 (Perfect Score)
Strong foundation in Mathematics, Physics, Chemistry, and Biology
Featured Research Projects
Comprehensive AI-driven platform that automates question generation and script evaluation using Large Language Models. The system employs semantic-aware grading to provide immediate, unbiased feedback, significantly reducing teacher workload while enabling students to receive consistent, meaningful assessment that promotes self-learning.
Key Innovations: Semantic evaluation engine, automated question bank generation, handwritten answer processing, curriculum-aligned assessment, multi-format input support (text, images, PDFs)
Developed attention-based neural network model for Bengali text summarization, addressing challenges in low-resource language processing. Implemented novel tokenization and attention mechanisms for long article compression.
Machine learning platform for stock market analysis and prediction. Implemented time series analysis using LSTM and Transformer models in short-term price prediction with comprehensive portfolio optimization features.
Advanced chess engine implementing minimax algorithm with alpha-beta pruning and position evaluation functions. Capable of playing at 1400+ ELO level with multiple difficulty settings and move analysis features.
Blockchain-inspired payment platform ensuring transaction transparency for NGOs. Features multi-platform support, secure payment integration, and real-time transaction tracking to increase donor confidence and organizational accountability.
Honors & Recognition
Deep Learning competition recognizing excellence in machine learning and AI research
Teams: DU_TripleHash, DU_Fireflies | Advanced to finals of Bangladesh's premier hackathon, 3rd in Phase-2
Team: DU_StormSurge | National Java programming and software development competition
Demonstrated excellence in scientific problem-solving and analytical thinking at national level
Academic Service
- Lead organization of academic conferences, workshops, and technical seminars for 300+ students
- Taken the initiative and organized debating competitions from 2022 to 2024
- Organized CTF contest for the first time in the Department of Computer Science and Engineering, University of Dhaka
- Organize guest lectures featuring prominent researchers and industry professionals
Maintained official records, coordinated inter-departmental activities, organized study groups, and technical skill development.
Contributed to event organization, student welfare initiatives, and department-wide activities supporting academic and extracurricular excellence.