Salam Thabet Doghmash

Deep Learning Transformers NLP

Building deep-learning systems that listen across languages — at production scale.

01 — Profile

About me ·

I am a PhD researcher in Computer Science at Gazi University, specializing in deep learning, Natural Language Processing, and multilingual hate speech detection. With 4+ years in applied machine learning and 6+ years in software engineering, I bridge rigorous research with production-ready engineering.

My current research evaluates transformer architectures (XLM-RoBERTa, MARBERT) for cross-lingual content moderation in Arabic, Turkish, and English — with published work at ICE 2025 and ongoing PhD research at Gazi University.

02 — Research

Research & Publications

2026
JOURNAL

Android application for multilingual hate speech detection: using XLM-RoBERTa and ONNX optimized

Journal of Computer & Electrical and Electronics Engineering Sciences (JCEEES) · Manuscript №29363 · S. T. Doghmash, İ. A. Doğru, K. Kılıç

Implements a real-time multilingual hate-speech detection system as a native Android app. XLM-RoBERTa is fine-tuned on Arabic, Turkish, and English data, then exported and quantised to ONNX for on-device inference. Demonstrates the practical deployment of transformer models to mobile environments with optimised latency and memory footprint.

2025
CONFERENCE

Multilingual Hate Speech Detection with XLM-RoBERTa

International Conference of Innovative Computer Engineering (ICE 2025) · S. Doghmash et al.

Designed and evaluated a multilingual ML pipeline for hate speech detection in Arabic, Turkish, and English. Implemented baseline models (TF-IDF + LR, SVM, RF) using a soft-voting ensemble, then fine-tuned XLM-RoBERTa with GPU acceleration, balanced sampling, and early stopping. Achieved 72.3% accuracy and 0.7079 Macro F1. Models were exported to ONNX and evaluated for real-time inference via API and Android deployment.

2022
THESIS

Detecting and Cleaning Arabic Hate Speech Texts Using Deep Learning Approach

Islamic University of Gaza · MSc Thesis · S. Doghmash (Supervisor: Dr. Motaz Saad)

Master's thesis addressing two problems: Arabic hate speech detection using deep learning models (RNN, CNN, and Transformers); and cleaning offensive/hate speech texts using a machine translation model with a parallel corpus of clean and dirty sentences. Includes analysis of dialectal variation, annotation noise, and dataset bias.

— Crafted on the web

Selected work

A selection of websites, dashboards, and web systems I have shipped over 7+ years — for charities, government, agencies, and product companies.

Mayor Files Archive — Ministry of Local Government

Internal System 2020

Ministry of Local Government, Gaza

Built an internal archival system for the mayor's office to digitize and search physical documents — using CodeIgniter PHP framework with Oracle database.

CodeIgniter PHP Oracle jQuery

Sirius for Information Technology

Web App 2021

Sirius — Gaza

Project Manager for multiple websites and mobile application projects. Owned planning, requirements analysis, and team coordination across Agile and Waterfall workflows.

Agile Waterfall Slack Trello Smartsheet

Coktilat — Charity Websites Suite

WordPress 2019

Coktilat

Developed and shipped multiple WordPress websites for charitable organizations, with custom themes, multilingual content, and donation flows.

WordPress PHP MySQL ACF Custom Plugins
06 — Available for client work

Build with me

I take on a small number of client projects each quarter — building multilingual websites, custom web systems, and AI-powered applications.

Websites

Corporate, community, and event sites — built to be fast, accessible, and multilingual from day one.

  • Multilingual · Arabic / English / Türkçe
  • Accessibility (WCAG-AA) and SEO ready
  • Optional admin panel for content updates

AI Integration

Bring deep-learning models into production: text classification, multilingual NLP, content moderation, recommendation systems.

  • PyTorch · Transformers · ONNX
  • On-device or cloud inference
  • Practical mobile (Android) and web pipelines
Message me on LinkedIn

Booking client projects for the next quarter. I reply within 48 hours.

04 — Experience

Professional journey

Oct 2022 — Present Present

PhD Researcher — AI & NLP

Gazi University

Ankara, Turkey

Researching multilingual transformer architectures, model evaluation methodologies, reproducibility, and ethical AI. Published at ICE 2025 with ongoing thesis work in cross-lingual content moderation.

Mar 2023 — Mar 2024

Project Manager & Web Developer

LINKON

Remote

Led cross-functional technical teams delivering data-driven web and analytics solutions. Managed full project lifecycle using Agile and Waterfall methodologies.

Jan 2021 — Dec 2022

Data Scientist

AutoECG · UCAS Incubator

Gaza, Palestine

Developed ML models for ECG signal analysis and classification. Led preprocessing, feature extraction, and supervised model training in an applied AI healthcare project.

Sep 2021 — Aug 2022

Lecturer

Al-Israa University

Gaza, Palestine

Taught Web Programming, Databases, and CMS courses. Supervised student projects and provided academic mentoring.

Jan 2018 — Jan 2023

Web Developer / Project Manager

Various Organizations

Gaza & Remote

Designed and developed multiple websites and systems for public and private organizations. Led teams from planning to delivery.

Sep 2017 — Jun 2018

Teaching Assistant

Islamic University of Gaza

Gaza, Palestine

Assisted in teaching programming and IT courses. Supported students in labs and project development.

Jul 2017 — Dec 2019

Web Development Trainer

Al-Raed Center

Gaza, Palestine

Delivered hands-on training in HTML, CSS, JavaScript, Bootstrap, and PHP — helping trainees move from fundamentals to real projects.

05 — Academic background

Education

2022 — Present

PhD

Computer Science

Gazi University

Research focus on multilingual NLP, transformer-based hate speech detection, model evaluation, and reproducibility.

2018 — 2022

MSc

Information Technology — Data Science

Islamic University of Gaza

Thesis: "Detecting and Cleaning Arabic Hate Speech Texts Using Deep Learning Approach". Built models with RNN, CNN, and Transformers, plus a parallel corpus for clean/dirty text translation.

Supervisor: Dr. Motaz Saad

2013 — 2017

BSc

Information Technology

Islamic University of Gaza

Foundation in software engineering, databases, web technologies, and computing fundamentals.

07 — Contact

Let's work together

Open to research collaborations, ML engineering roles, and consulting opportunities. The fastest way to reach me is by email or LinkedIn.