IEEE Access 2025
BERT-OTA: Ontology-Guided Transformer Architecture
Dual-stream architecture integrating BERT with Graph Neural Networks and scaled attention for binary hate speech detection. Achieved 91.30% accuracy, establishing a new benchmark.
Researcher & Data Scientist
Mahmoud Abusaqer
Specializing in NLP, LLMs, deep learning, and transformer architectures. Author of 9+ publications in venues including IEEE Access, ICMLA, and SEKE.
About
I am a researcher and engineer working at the intersection of natural language processing, large language models, and deep learning. My work focuses on designing architectures that combine the contextual power of transformers with structured knowledge from graph neural networks, resulting in 9 publications in venues such as IEEE Access, JVLC, ICMLA, SEKE, and ACMSE.
I am broadly driven by how AI systems learn, reason, and generalize. Whether benchmarking 303 model configurations for hate speech detection or evaluating parameter-efficient fine-tuning across 20 LLM architectures, I thrive on rigorous experimentation and collaborative, interdisciplinary problem-solving.
On the industry side, I have worked as a Software and Data Science Engineer at Metachain Technologies and as a Software Engineer at Xenomics, where I designed scalable backend systems and integrated machine learning models into production applications.
Research Interests
Natural Language ProcessingLarge Language ModelsDeep LearningTransformer ArchitecturesGraph Neural NetworksComputer Vision
Experience
Research, industry, and teaching roles
May 2024 – Present
Graduate Research & Teaching Assistant
Missouri State University · Springfield, MO
- Developed novel neural architectures combining transformer-based models with graph neural networks for NLP tasks.
- Conducted comprehensive evaluations of large language models, including parameter-efficient fine-tuning techniques.
- Published 7 papers in peer-reviewed journals (IEEE Access, JVLC) and conferences (CSCI, ICMLA, SEKE, ACMSE).
- Assisted instruction in Data Structures, Database Systems, and Data Science courses.
PyTorchTransformersLLMsLoRA/PEFTGNNs
Feb 2023 – Apr 2024
Software & Data Science Engineer
Metachain Technologies Inc.
- Led the integration of AI and AR in Unity projects, implementing ML models that enhanced user engagement.
- Developed scalable backend services using Flask and Django frameworks on cloud infrastructure.
- Worked on standalone data science projects, implementing predictive models and analytics solutions.
PythonFlaskDjangoGCPUnityMachine Learning
Aug 2022 – Feb 2023
Software Engineer
Xenomics
- Developed scalable software solutions using Python, Django, and React.js for enterprise applications.
- Designed and implemented RESTful APIs following MVC design patterns, ensuring high performance and code quality.
- Contributed to the entire software development lifecycle, from requirements analysis to deployment.
PythonDjangoReact.jsPostgreSQLREST APIs
Aug 2022 – Aug 2023
Teaching Assistant
Islamic University of Gaza · Gaza, Palestine
- Led instruction for Programming 1, 2, and 3 lab sessions; developed hands-on exercises reinforcing theory.
- Achieved measurably better student outcomes: higher grades, lower dropout rates compared to other sections.
JavaC++TeachingData Structures
Projects
Research and engineering projects
CSCI 2025
RoBERTa-OTA: Multiclass Hate Speech Detection
Dual-stream architecture combining RoBERTa with Enhanced 3-Layer GNN for 5-class hate speech detection. Achieved state-of-the-art 96.06% F1-score, outperforming previous best by 1.62 percentage points.
JVLC & SEKE 2025
LLM Hate Speech Detection Comparative Study
Evaluated 20 LoRA-adapted LLMs (Llama, Phi, Qwen) and three zero-shot API models for hate speech detection. Best fine-tuning achieved 91.84% accuracy with only 0.15% trainable parameters.
Agentic AI University Advisor
Autonomous AI agent using Claude API for university course registration assistance. Implements agentic architecture with web scraping pipeline, Flask, and context-aware conversation management.
Claude APIFlaskAgentic AIPython
SpeakGPT – AI Voice Assistant
Cross-platform mobile AI assistant using Unity and C# with OpenAI API integration. Real-time NLP with text-to-speech and speech-to-text capabilities for multiple languages.
UnityC#OpenAI APITTS/STT
Automated University Timetabling System
Genetic algorithm-based optimization system generating conflict-free schedules for 2000+ students across 38+ departments with custom mutation operators.
PythonGenetic AlgorithmOptimization
Publications
9 papers across IEEE, ACM, and other top venues
Conference
Comparative Analysis of Boosting Algorithms and DistilBERT Across Multi-Domain Text Classification Tasks
M. Abusaqer, S.M. Faiaz Mursalin, and J. Saquer
ACM Southeast Conference (ACMSE), 2026 (submitted)
Conference
Comparing Transformers and Zero-Shot Large Language Models for Suicidal Ideation Detection
M. Abusaqer, J. Nwobodo, K. Hasan, and J. Saquer
ACM Southeast Conference (ACMSE), 2026 (submitted)
Conference
Multiclass Hate Speech Detection with RoBERTa-OTA: Integrating Transformer Attention and Graph Convolutional Networks
M. Abusaqer, J. Saquer
International Conference on Computational Science and Computational Intelligence (CSCI), 2025
Peer Reviewer -- IEEE Access (5 verified reviews) | ORCID: 0009-0006-2850-6566
Technical Skills
Programming Languages
PythonJavaC++C#JavaScriptSQL
Machine Learning & AI
PyTorchTensorFlowHugging Face Transformersscikit-learnBERT / RoBERTaLLMs (Llama, Phi, Qwen)LoRA / PEFTRAG
NLP & Deep Learning
Text MiningSentiment AnalysisspaCyNLTKGraph Neural NetworksCNNsComputer VisionOpenCV
Data Science
pandasNumPyMatplotlibSeabornPlotlyStatistical AnalysisData MiningPredictive Modeling
Software Engineering
FlaskDjangoSpring BootReact.jsNode.jsREST APIsGitAgile / Scrum
Databases & Cloud
MySQLPostgreSQLMongoDBOracleGCPAWSDockerLinux
Education
Master of Science in Computer Science
Missouri State University · Springfield, MO
Jan 2024 – May 2026 (Expected)
GPA: 3.76 / 4.0
Thesis
Advancing Hate Speech Detection: A Comprehensive Evaluation of Traditional, Deep Learning, and Parameter-Efficient Transformer Approaches with Ontological Knowledge Integration
Machine LearningDeep LearningComputer VisionAdvanced Data AnalyticsData MiningResearch Methods
Bachelor of Science in Computer Science
Islamic University of Gaza · Gaza, Palestine
Sep 2018 – Aug 2022
GPA: 86 / 100
Ranked 4th in graduating class
Senior Project
Automated University Timetable Scheduling Using Evolutionary Algorithms with Shotgun Hill-Climbing Optimization
Awards & Certifications
Outstanding Poster Presentation Award
Einhellig Interdisciplinary Forum (EIDF), Missouri State University
Graduate Assistantship
Missouri State University
Ranked 4th in Class
Bachelor of Computer Science, Islamic University of Gaza
Certifications
AI Agents Course Certificate
Hugging Face
DataCamp Professional Data Scientist
DataCamp
ML Bootcamp
Gaza Sky Geeks
IEEE Access Peer Reviewer
5 verified reviews
Get in Touch
Open to research collaborations, industry positions, and new opportunities
Find Me Online
Open to relocation. U.S. work authorization in progress.