Built a multi-class text classification model to categorize Arabic news articles using LSTM. The pipeline included advanced preprocessing (diacritics removal, tokenization, padding), class balancing, and model training with early stopping. Achieved strong performance with 73% accuracy showing effectiveness in handling imbalanced text data.

Built a CNN model to accurately distinguish between dogs and cats, aiming to reduce misclassification in noisy image datasets. Applied median filtering to clean salt-and-pepper noise and used data augmentation to improve generalization. Achieved 85.7% validation accuracy, solving the challenge of reliable classification in real-world.

This project simulates real-time robot control in a browser using HTML, CSS, JavaScript, and PHP. Movements are visualized on a canvas and stored in a PHPMyAdmin database via AJAX. Built to solve the need for testing movement logic interactively, it enables live feedback, persistent command logging, and a user-friendly interface for demos or learning.

I designed this bracket using Tinkercad to enable precise vertical and horizontal head movement in a robot. It solves the challenge of limited motion by using the MG996R servo motor, allowing the robot to scan, track, and interact more effectively with its surroundings improving responsiveness and sensor alignment.

Built a distance-aware system using an HC-SR04 sensor and Arduino Uno to detect nearby objects and trigger an LED. The goal was to create a simple proximity alert solution. It helps automate responses based on distance useful in obstacle detection or safety systems while reinforcing circuit design and sensor integration fundamentals.