Abdullah Ibne Hanif Arean

• Selected as one of 26 students and researchers from universities in Bangladesh and abroad for the ICT Division Research Fellowship under the 2025–2026 fiscal year.
• Awarded the fellowship following a competitive national selection process involving research proposal screening, document verification, interview, and presentation before the selection and award committees.

• Selected as one of 48 fellows for a 16-week AI safety research fellowship involving 12 mentors and 12 research projects.
• Contributing 20+ hours per week to a structured research program focused on methodology selection, baseline development, active experimentation, paper drafting, and final research release.
• Working through mentor-led quality gates and peer-driven research sprints to develop a submission-ready AI safety paper for fall conference deadlines.

• Developing graph-based pipelines and chain-of-thought reasoning to improve multimodal interpretation in Vision-Language Models (VLMs).
• Leading research on specialized VLMs for diagram understanding, addressing dataset and modeling limitations.
• Integrating explainable AI principles to address black box problems and enable deployment in safety-critical domains.

• Advising on digital transformation initiatives for EIIN-registered schools and colleges through Amar IMS, focusing on academic and administrative automation.
• Supporting SaaS-based business solutions for data collection, payment management, and operational efficiency.
• Providing consulting and academic advising to help deploy scalable, secure, and cost-effective IT solutions.

• Co-leading the development of end-to-end AI systems designed to solve real enterprise problems with measurable business impact.
• Overseeing production-ready AI deployment pipelines for scalable enterprise integration.
• Working on AI, NLP, and research-backed intelligent systems to improve business automation and decision-making.

• Collected multiple datasets in Bangla language spanning more than 50 million Bangla sentences.
• Built multiple LLM-based models for Bangla summarization, translation, and other tasks.
• Worked on creating handwritten image recognition for Bangla.

• Worked on developing AI-driven tools using advanced techniques including large language models.
• Collaborated with engineers on automation, validation, and verification workflows.

• Worked on AI implementation in fish production.
• Collaborated with the team to integrate AI solutions in real-time systems.

• Conducted interviews and system designing with stakeholders (teachers) to develop a system automating the Masters admission program at the University of Dhaka from scratch.
• Worked on backend and database development.
• Collaborated in a team to complete the project within the stipulated timeframe and maintained it throughout its lifecycle.

• Worked on custom Linux kernel development for a storage device.
• Contributed to system-level programming in C to support low-level storage device functionality.
• Focused on improving storage device performance through kernel customization and system-level optimization.

Stroke-Level Connectivity Verification: Grounding Vision-Language Models Against Topology Hallucination

20th International Conference on Document Analysis and Recognition 2026 | Acceptance, Full PDF

Authors: Abdullah Ibne Hanif Arean, Niamul Hassan Samin, Md Arifur Rahman, Renu Akter Suity, Juena Ahmed Noshin, Md Ashikur Rahman

• Proposes Stroke-Level Connectivity Verification (SLCV), enforcing pixel-grounded proof of connectivity via continuous stroke paths in diagrams.
• Introduces PaRCO, a modular architecture separating structural reconstruction, semantic reasoning, and constraint validation.
• Establishes topology-aware verification as a reusable primitive for multimodal document understanding tasks.

SOMAJGYAAN: Evaluating LLMs on Bangla Culture and Social Knowledge

AACL-IJCNLP 2025 Findings

Authors: Fariha Anjum Shifa, Muhtasim Ibteda Shochcho, Abdullah Ibne Hanif Arean, Mohammad Ashfaq Ur Rahman, Akm Moshiur Rahman Mazumder, Ahaj Faiak, Md Fahim, M Ashraful Amin, Amin Ahsan Ali, AKMMAHBUBUR RAHMAN

• Introduces a 4,234-question Bangla multiple-choice benchmark; LoRA fine-tuning narrows gaps for open models but gaps remain.
• Highlights non-STEM cultural and social knowledge coverage unique among existing Bangla datasets.

Automatic Vehicle Detection Using DETR: A Transformer-Based Approach for Navigating Treacherous Roads

arXiv Preprint, Feb 2025

Authors: Istiaq Ahmed Fahad, Abdullah Ibne Hanif Arean, Nazmus Sakib Ahmed, Mahmudul Hasan

• Proposed one of the first DETR-based frameworks for automatic vehicle detection in challenging driving environments, leveraging Co-DETR to improve supervision and feature learning.
• Evaluated multiple DETR variants and YOLO baselines on the BadODD dataset, demonstrating improved robustness under diverse road, lighting, and weather conditions.
• Introduced collaborative hybrid assignment training with auxiliary heads to enhance training efficiency and detection performance for real-world autonomous navigation systems.

• Engineered a web-based news aggregation platform that ingests, normalizes, and summarizes Bangla articles from multiple publishers in real time.
• Designed and implemented LongBanSum, a T5-based transformer with combined local–global attention (TGlobal) optimized for long-sequence Bangla summarization.
• Curated the MASBA corpus of 669,699 articles with multi-level human-written summaries, ensuring diverse, high-quality data for model training.

• Collaborated on full-stack development of a centralized portal delivering on-demand IT support and solutions for common technical issues.
• Led design of service workflows, created technical documentation, and maintained release cycles.
• Implemented core services using Spring Boot, Spring MVC, and Java, integrating authentication and RESTful APIs.

• Developed an Android application in Java to streamline disaster relief distribution and resource tracking in affected regions.
• Designed a secure backend API and SQLite database schema to ensure data integrity, real-time updates, and offline resilience.

• Led end-to-end development of a scalable attendance management system, coordinating stakeholder interviews and requirement elicitation.
• Implemented biometric and QR-code check-in features with JavaFX frontend, ensuring seamless user experience and data accuracy.

• Architected a modular hospital management system using JavaFX for the UI and MySQL for backend operations.
• Developed patient registration, appointment scheduling, and inventory modules with secure role-based access control.

• Created a C/C++ action game using SDL2 to honor Bangladeshi Freedom Fighters with historically inspired levels and narratives.
• Implemented custom physics, sprite animations, and AI-driven enemy behaviors for an engaging gameplay experience.
• Designed sound effects and a dynamic scoring system, and packaged builds for Windows and Linux distributions.

• Conducted sessions on PC building and basic computing, mastering Google search for academic excellence, and cybersecurity awareness for women.
• Contributed actively to DUITS' mission of promoting IT education, organizing events, workshops, and fostering collaboration among students and professionals.
• Engaged in community-building, networking, and sharing IT expertise within the society.

• Collected and analyzed data on casualties and martyrs from the July Revolution, contributing to historical records and policy development collaborating with MIS.
• Student representative of Shaheed Suhrawardy Medical College and led the Help Center of the July Shaheed Smrity Foundation.

• Leading the IEEE Computer Society Student Branch Executive Committee (Excom) at Dhaka University for the year 2024.
• Coordinating with the full committee to organize events, workshops, and seminars focused on science and technology.
• Promoting IEEE Computer Society initiatives and fostering collaboration among students and professionals.