All Projects
ML Wiki RAG Pipeline
End-to-end Retrieval-Augmented Generation system with multiple chunking strategies (fixed, sentence, semantic, recursive), hybrid dense + sparse retrieval with cross-encoder re-ranking, and comprehensive IR evaluation metrics.
ChessGPT
GPT-style transformer trained to predict chess moves by learning from a grandmaster's games. Built a custom tokenizer for chess notation, implemented logit masking for legal move generation, and explored how language models can capture individual playing styles.
Surah Quiz
Interactive quiz game where users select a surah and are challenged to fill in the blank for missing words in each ayah. Tests knowledge of the Quran through an engaging, educational format.
Interactive Tic-Tac-Toe with AI
Web-based tic-tac-toe game featuring 5 AI opponents implementing different strategies: Random, Goal-Based, Utility-Based evaluation, Minimax, and Alpha-Beta Pruning. Explored game theory algorithms and optimal decision-making in adversarial games.
Make Happy - Mental Wellness
A user-centered web application designed to promote mental wellness by generating whimsical, humorous, and playful personalized sentences about users. Built with modern React patterns and Next.js, features dynamic content generation and responsive design principles.
Islamic Flashcard Generator
This tool converts Islamic book PDFs into Quizlet-compatible flashcards. Features automatic chapter/section detection, AI-powered learning point extraction, Arabic text preservation, and bilingual support (Arabic/English).
Basic LLM from Scratch
Built a large language model from the ground up, implementing transformer architecture, attention mechanisms, and training pipeline. Demonstrates deep understanding of neural network fundamentals and NLP.
ML for Space Debris Orbital Paths
Machine learning model to predict and analyze orbital trajectories of space debris. Applied predictive modeling techniques to address space sustainability challenges.
Middle-Earth Messenger
A Streamlit chat application that lets you converse with characters from The Lord of the Rings. Each character speaks in their authentic voice, drawing from their quotes and wiki lore. Features character selection, conversation memory, and rich lore integration.
🎲 Ludo Master
A Ludo game built with Phaser.js as a Progressive Web App (PWA). Features classic gameplay with smooth animations, local multiplayer support (2-4 players), animated dice, safe spots, and victory celebrations. Works on web browsers, Android, and iOS devices with offline support.
MNIST & Iris Dataset Analysis
Comprehensive data analysis and machine learning exploration of classic datasets. Implemented various classification algorithms, feature engineering, and visualization techniques to extract insights.