Mosab Rezaei
Mosab Rezaei
Embedded Files
I’m a Ph.D. candidate in Computer Science with over 10 years of experience in Artificial Intelligence (AI), both theoretical and applied, specializing in deep learning, natural language processing (NLP), natural language generation (NLG), and large language models (LLMs). I build end-to-end machine learning systems, from optimization and deployment to explainability, with hands-on work in computer vision and object detection. My current research focuses on stylometry and style-conditioned text generation, combining classical machine learning methods with deep neural networks to improve efficiency, interpretability, and adaptability of LLMs. I’m also designing and researching AI agents with reasoning strategies, tool use, and RAG.
Education
Education
Aug 2020 - Present:
Ph.D. Candidate in Computer Science
Northern Illinois University, DeKalb, Illinois
Dissertation: "Generation, Evaluation, and Explanation in the Writing Style of Different Authors"
Aug 2015 - Feb 2018:
M.S. in Artificial Intelligence and Robotics
Shahid Chamran University of Ahvaz, Ahvaz
Thesis: "Face Detection Using Deep Neural Networks"
Research Interests
Research Interests
Stylometry
AI Agent
Deep Learning
Object Detection
Face Recognition
Large Language Models (LLM)
Natural Language Understanding (NLU)
Natural Language Processing (NLP)
Natural Language Generation (NLG)
Experiences
Experiences
Aug 2020 - Present:
Teaching Assistant
Department of Computer Science, Northern Illinois University
Jul 2018 - Feb 2020:
Research Assistant
ITML Laboratory, Shahid Chamran University of Ahvaz
Mar 2018 - Jun 2018:
Teaching Assistant
ITML Laboratory, Shahid Chamran University of Ahvaz
2025:
Detecting, Generating, and Evaluating in the Writing Style of Different Authors [Link]
Mosab Rezaei.
Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL) in Student Research Workshop (SRW)
2025:
Differentiating U.S. Presidential Written Texts from Transcribed Speeches Using Word Dependencies in Graph Neural Networks [Link]
Mosab Rezaei, Miguel Williams, Reva Freedman, and Lei Zhang.
Midwest Speech and Language Days (MSLD)
2024:
Text vs. Transcription: A Study of Differences Between the Writing and Speeches of US Presidents [Link]
Mina Rajaei Moghadam, Mosab Rezaei, Gülşat Aygen, and Reva Freedman.
4th International Conference on Natural Language Processing for Digital Humanities (NLP4DH)
2024:
Investigating Lexical and Syntactic Differences in Written and Spoken English Corpora [Link]
Mina Rajaei Moghadam, Mosab Rezaei, Miguel Williams, Gülşat Aygen, and Reva Freedman.
37th International Florida Artificial Intelligence Research Society Conference (FLAIRS)
2023:
Test Case Recommendations with Distributed Representation of Code Syntactic Features [Link]
Mosab Rezaei, Hamed Alhoori, and Mona Rahimi.
38th IEEE/ACM International Conference on Automated Software Engineering Workshops (ASEW)
2019:
Assessing the Effect of Image Quality on SSD and Faster R-CNN Networks for Face Detection [Link]
Mosab Rezaei, Elhamossadat Ravanbakhsh, Ehsan Namjoo, and Mohammad Haghighat.
27th Iranian Conference on Electrical Engineering (ICEE)
2016:
Comparison Between EM and FCM Algorithms in Skin Tone Extraction [Link]
Elham Ravanbakhsh, Ehsan Namjoo, Mosab Rezaei, and Padideh Choobdar.
1st International Conference on New Research Achievements in Electrical and Computer Engineering (ICNRAECE)
Talks
Talks
25 Apr 2025:
Invited speaker for the AI/ML Seminar on “LoRA, QLoRA, DoRA: Low-Rank Adaptation of Large Language Models” [Link]
Northern Illinois University
7 Mar 2025:
Invited speaker for the AI/ML Seminar on “Forward-Forward Algorithm” [Link]
Northern Illinois University
29 Jan 2025:
Invited speaker for the panel on “AI in Industry and Research” [Link]
National Autonomous University of Mexico (IIMAS-UNAM)
18 Dec 2019:
Workshop Instructor, “Introduction to Deep Learning in Medical Science”
Department of Biostatistics, Kermanshah University of Medical Sciences (KUMS)
Honors
Honors
2024:
Member of Phi Beta Delta: Honor Society for International Scholars
Northern Illinois University, DeKalb, IL
Deep Learning - NLU, NLP, NLG:
Stylometry: Designed an LLM framework for style-conditioned text generation; improved efficiency with LoRA fine-tuning techniques on "GPT 3-Neo 1.3B"; added explainability via a DeBERTa-based classifier, Attention Enrichment, Integrated Gradients, and Traditional ML models. Published at JCDL & NAACL-SRW.
Writing & Speech Analysis: Performed corpus-level lexical and syntactic analysis using Stanford CoreNLP and NLTK; extracted low- and high-level features to characterize class-distribution patterns and their impact on ML classifiers (decision trees, random forests, SVMs, BERT). Results published at MSLD, FLAIRS, and NLP4DH.
Graph Neural Networks: Built and evaluated GNNs to classify written vs. spoken sentences by U.S. presidents and to distinguish 19th-century novelists’ styles. Implemented message-passing (GraphSAGE), attention-based (GAT), spectral (GCN, ChebNet), and expressive (GINConv) architectures. Results presented at MSLD.
ML\DL:
Face Recognition: Developed joint face-detection and gender-recognition models using Faster R-CNN and SSD on the WIDER FACE and IMDb datasets to evaluate DNN robustness to image noise and compression; results published at ICEE.
Face Detection: Implemented shallow CNN classifiers trained on the WIDER FACE dataset to detect faces with different sizes in a crowded image. Results reported in Maser’s thesis.
Face Skin Detection: Applied Expectation-Maximization and Fuzzy C-Means clustering for robust skin-region segmentation in complex images.
Source Code Recommendation: Embedded code snippets with shallow neural networks using structural and semantic features to recommend the best test cases for target methods.
Genetic Algorithms: Developed a Java application that finds the shortest (near-optimal) paths on arbitrary maps using a genetic algorithm.
Data Visualization:
NFL Player Performance: Developed an interactive web visualization in Observable (JavaScript) for the NFL dataset to compare player performance and abilities.
Threats in Textual Data: Developed an interactive web visualization to identify and prioritize prominent threats across textual police reports.
Software - Gaming & Simulation:
Christopher Columbus Game: Implemented a graphical Java game applying software-engineering design patterns like Singleton, Observer, Strategy, Composite, and Decorator.
Computer Simulation: Implemented a C++ simulator capable of loading and executing machine-language programs.
Professional Service
Professional Service
2025
Program Committee (Reviewer)
JCDL 2025, ACM/IEEE Joint Conference on Digital Libraries
Page updated
Google Sites
Report abuse