Ph.D. Candidate, Department of Computer Science, Northern Illinois University
I have experience working in the IT industry for over a decade. I completed my bachelor’s degree in Computer Science from the University of Dhaka and my master’s in Computer Science from Northern Illinois University. I am pursuing my Ph.D. in Computer Science at Northern Illinois University.
I am proficient in working with Machine Learning (ML) models with advanced knowledge in Computer Vision (CV), Natural Language Processing (NLP), Artificial Intelligence (AI), Human-Computer Interaction (HCI), and Information Visualization (InfoViz), evident in several relevant peer-reviewed publications. With my expertise in Python libraries (e.g., Spacy, TensorFlow, PyTorch, etc.) and JavaScript tools (e.g., D3.js), I am able to extend my work in different directions. My recent publication expands research boundaries in non-destructive inspection (NDI), proposing image segmentation models to identify defects from ultrasonic and CT scans. I am currently working on multimodal LLMs for ROI detection.
Expertise
- Image Segmentation
- Data Visualization
- Machine Learning
- Artificial Intelligence
- Data Analytics
- Natural Language Processing
Research Interests
- Machine Learning
- Artificial Intelligence
- Computer Vision
- Visual Analytics
- Social Media
- Human-Computer Interaction
Publications
Proposing a Framework for Machine Learning Adoption on Legacy Systems
The integration of machine learning (ML) is critical for industrial competitiveness, yet its adoption is frequently stalled by the prohibitive costs and operational disruptions of upgrading legacy systems. The financial and logistical overhead required to support the full ML lifecycle presents a formidable barrier to widespread implementation, particularly for small and medium-sized enterprises. This paper introduces a pragmatic, API-based framework designed to overcome these challenges by strategically decoupling the ML model lifecycle from the production environment. Our solution delivers the analytical power of ML to domain experts through a lightweight, browser-based interface, eliminating the need for local hardware upgrades and ensuring model maintenance can occur with zero production downtime. This human-in-the-loop approach empowers experts with interactive control over model parameters, fostering trust and facilitating seamless integration into existing workflows. By mitigating the primary financial and operational risks, this framework offers a scalable and accessible pathway to enhance production quality and safety, thereby strengthening the competitive advantage of the manufacturing sector.
VulSim: Leveraging Similarity of Multi-Dimensional Neighbor Embeddings for Vulnerability Detection
Natural Language Processing Journal
Despite decades of research in vulnerability detection, vulnerabilities in source code remain a growing problem, and more effective techniques are needed in this domain. To enhance software vulnerability detection, in this paper, we first show that various vulnerability classes in the C programming language share common characteristics, encompassing semantic, contextual, and syntactic properties. We then leverage this knowledge to enhance the learning process of Deep Learning (DL) models for vulnerability detection when only sparse data is available.
Effective Defect Detection Using Instance Segmentation for NDI
Ultrasonic testing is a common Non-Destructive Inspection (NDI) method used in aerospace manufacturing. However, the complexity and size of the ultrasonic scans make it challenging to identify defects through visual inspection or machine learning models. Using computer vision techniques to identify defects from ultrasonic scans is an evolving research area. In this study, we used instance segmentation to identify the presence of defects in the ultrasonic scan images of composite panels that are representative of real components manufactured in aerospace.
Mixture-of-Experts for Multi-Domain Defect Identification in Non-Destructive Inspection
Despite their effectiveness, traditional UT methods rely on the manual interpretation of ultrasonic signals, which is time-consuming, labor-intensive, and subjective. Furthermore, processing such large-scale data, particularly across materials of varying thicknesses, significantly increases the computational demands of deep learning model optimization. To overcome these challenges, we propose an efficient sparse mixture-of-experts (MoE) model with a multi-level loss function and introduce four novel training objectives to improve computational efficiency and accuracy in identifying surface defects in composite aircraft materials.
Cutting through the noise to motivate people: A comprehensive analysis of COVID-19 social media posts de/motivating vaccination
Natural Language Processing Journal
The COVID-19 pandemic exposed significant weaknesses in the healthcare information system. The overwhelming volume of misinformation on social media and other socioeconomic factors created extraordinary challenges to motivate people to take proper precautions and get vaccinated. In this context, our work explored a novel direction by analyzing an extensive dataset collected over two years, identifying the topics de/motivating the public about COVID-19 vaccination.
Public interest in science or bots? Selective amplification of scientific articles on Twitter
ASLIB Journal of Information Management
Since public interest in scientific findings can shape the decisions of policymakers, it is essential to identify the possibility of bot activity in the dissemination of any given scholarly article. Without arguing whether the social bots are good or bad and without arguing about the validity of a scholarly article, our work proposes a tool to interpret the public interest in an article by identifying the possibility of bot activity toward an article.
Visual Analysis of COVID-19 Vaccine Stance
We developed a web-based tool to analyze COVID-19 tweets visually. This interactive tool helps identify relations and hidden patterns in COVID-19 vaccine-related topics and public stance towards vaccination.
Analyzing Twitter Bot Activity on Academic Articles
Social Media & Society Conference 2020
A short paper analyzing the Twitter bot activity on academic articles was published at the Social Media & Society conference 2020. The co-authors and I analyzed the behavior of Twitter bots using the dataset from Altmetrics.
Web Search Engine Misinformation Notifier Extension (SEMiNExt): A Machine Learning Based Approach during COVID-19 Pandemic
A research paper published in the MDPI – Healthcare journal where me and the co-authors proposed a machine learning based approach to catch misinformation related search keywords and add a roadblock before redirecting users to possible harmful healthcare behaviors.
Current Projects
Identifying Regions of Interest in CT Scans
I am developing a multimodal LLM framework to identify objects and regions of interest (ROI) from microscopic CT scans in scientific and industrial domains. The proposed framework uses an RAG-based approach in combination with Vision Language Models to identify the ROIs. Language-based reasoning and transparency of the decision-making are key focuses of the framework.
Education
Ph.D. in Computer Science – Northern Illinois University
Areas of research for the Ph.D. covers Computer Vision, Machine Learning, Multimodal LLM, NLP, visual analytics, and big data analytics.
2022 – running
M.Sc. in Computer Science – Northern Illinois University
Major topics covered in the graduate program were Python programming, big data analytics, machine learning, data visualization, and artificial intelligence.
2019 – 2022
B.Sc. in Computer Science – University of Dhaka
Major topics covered in the undergraduate program were computer programming, network security, database management, operating systems, discrete mathematics, and calculus.
2003 – 2009
Work Experience
Research Assistant – Northern Illinois University
Developed a CV pipeline for ultrasonic NDI using PyTorch, Detectron2, and YOLO, achieving above 80% mAP50 with minimal preprocessing, which is deployable on edge devices. Built a custom COCO-format dataset from 100+ annotated scans to address domain-specific data scarcity. Published at AI2ASE workshop, AAAI 2025.
Designed a standardized framework for ML adoption on legacy industrial systems, eliminating hardware upgrade requirements and lowering barriers to industry-wide deployment. Presented at RAiM workshop, ICDM 2025, Washington D.C.
Developed a model-free pseudo-labeling algorithm for microscopic ROI detection that runs on CPU-only resources while delivering 2× faster results than GPU-powered models like SAM. Currently extending this into a multimodal LLM pipeline for automated ROI identification in microscopic scans.
Applied NLP and data visualization to analyze large-scale social media datasets, identifying key motivating and deterring factors for COVID-19 vaccination. Published in the Natural Language Processing Journal and presented a visual exploration tool at JCDL 2023.
Built an NLP framework to detect spam and selective amplification of academic content on Twitter, achieving 70% detection accuracy across large-scale datasets. Published in the Aslib Journal of Information Management.
2022 – current
Visiting Student Researcher – Argonne National Laboratory
Worked with the Advanced Photon Source (APS) ML team at Argonne National Laboratory to develop unsupervised models, including CUTS and W-Net, for pseudo-label generation in complex CT scans, advancing automated ROI detection in scientific imaging workflows.
Fine-tuned zero-shot object detection models for image segmentation in specialized scientific imaging contexts, improving performance metrics by 15–30% over state-of-the-art foundation models.
Trained and evaluated models on the Lambda GPU cluster using distributed multi-GPU computing, optimizing workflows for large-scale scientific imaging datasets, including the APS Silica Sand dataset.
2025 – 2025
Teaching Assistant – Northern Illinois University
Mentored 200+ students across five CS courses including Software Engineering, C++, and UNIX, maintaining a 90% project completion rate while upholding industry-standard coding and documentation practices.
2019 – 2022
Web Application Developer – Choobs Ltd.
Developed and maintained an ML model for passenger flow prediction at Geneva Airport for easyJet, improving resource allocation and reducing boarding time by 30%.
Built and deployed ML-powered customer queue management models adopted by multiple businesses in the Geneva area, reducing average customer wait time by 80%.
Automated airline crew onboard documentation workflows, reducing operational delays by 20%.
Led cross-functional development teams on full-stack projects, including e-commerce and learning management systems, serving thousands of daily active users.
2011 – 2019
Founder & CEO – Creativity Unleashed
Founded and led a software company delivering automation solutions for 20+ organizations, including one of Bangladesh’s first e-commerce platforms.
2006 – 2013
Programmer – SGC Soft Ltd.
I worked on several desktop and web-based software solutions for small and medium sized organizations.
2009 – 2009
Achievements
- Travel Scholarship: AAAI (2025)
- Google Cloud Research Grand: Google (2023)
- SIGIR Student Travel Grant: JCDL (2023)
- Finalist: The Virus Versus Hackathon (2020)
- Participation: The AIR Hackathon (2018)
- Finalist: Porsche NEXT OI (2018)
- Finalist: Human Beyond Digital Hackathon (2018)
- Finalist: Hôpitaux Universitaires de Genève (HUG) – Hackathon (2018)
- Champion: Mercedes-Benz Digital Challenge (2017)
Activities
- Program Committee Member – WWW 2026
- Program Committee Member – JCDL 2025
- Program Committee Member – CIKM 2023, 2024, 2025
- Participated in STEM fest organized by NIU and present our work in NDI research (2022)
- Volunteered as a speaker at Inspiring Youth Bangladesh (2018)
- Participated in e-commerce Expo organized by the Bangladesh Association of Software and Information Services (BASIS) (2012)
- Participated at Soft Expo organized by Bangladesh Association of Software and Information Services (BASIS) as entrepreneur (2010, 2011, 2012)