Projects
A showcase of my work across machine learning, web development, and game design
Palestinian Dialect NLP
A text-to-text generation model trained on Palestinian Arabic dialect using literature and corpus data.
Workout Vision AI
Computer vision application that analyzes workout form and provides real-time feedback for improved training efficiency and safety.
Treadmill Network System
Comprehensive platform for managing networked treadmills with real-time data collection, race functionality, and multi-device synchronization.
Text Analysis Toolkit
Comprehensive suite of text analysis tools combining factor analysis, sentiment analysis, and machine learning categorization for processing survey and text data.
Music Downloader
A Python application that downloads music from various streaming platforms. Features include playlist synchronization, metadata preservation, and automatic organization.
Tower Strategy Game
A tower defense strategy game with multiple defensive towers, enemies, and resource management in an isometric game environment.
Castle Defender
A magical castle defense game where you control a wizard with a staff, casting spells to protect your castle from waves of enemies.
Kahoot Clone
A comprehensive Kahoot-like platform with a dashboard for controlling questions, pre/post tests, and real-time student data collection and analysis.
SafeScales
A comprehensive educational platform consisting of a gamified Flutter mobile app for social media literacy and safety education, paired with a React dashboard for educators to manage classes, field sites, and student progress.
Dagbani-English Neural Machine Translation
A neural machine translation system for Dagbani-English, a low-resource language pair. Built using fine-tuned mBART-large-50 with data collected from multiple sources including religious texts, dictionaries, poetry, and Wikipedia. Features parameter-efficient fine-tuning with LoRA and iterative model improvements achieving significant BLEU score gains (0.02 → 25.97). Model available on HuggingFace.
Cross-Cultural LLM Misunderstanding Analysis
A full-stack research tool for analyzing cross-cultural communication norm violations using Large Language Models. Evaluates how different LLM generations (GPT-4o, GPT-5, GPT-5.1, GPT-5.2) detect cultural misunderstandings and identifies relevant cultural dimensions. Features an automated experiment runner that generates prompts from demographic combinations and provides sophisticated visualization capabilities for comparative analysis across model versions.