Ahmad Qaraqish

Results-driven and fast-learning Artificial Intelligence and Data Science student in final year of study, with practical experience in machine learning, supervised learning, and audio recognition. Skilled in Python programming, TensorFlow, statistical modeling, data cleaning, feature engineering, and deep learning model development. Strong understanding of data preprocessing and model evaluation. Known for quick adaptability, analytical problem-solving, and a strong interest in advancing AI technologies. Eager to apply machine learning expertise to real-world applications in a collaborative and innovative environment.

Deep Learning Arabic Accent Recognition for Improving Transcription Accuracy (Graduation Project)

• Developed a deep learning model using a hybrid CNN-LSTM architecture to classify four Arabic dialects (Saudi, Khaliji, Egyptian, Standard) with ~85% accuracy.
• Designed a comprehensive preprocessing pipeline for Arabic audio data, including noise reduction, segmentation, and MFCC feature extraction.
• Evaluated model performance using metrics like precision, recall, F1-score, and ROC-AUC, with strong results for Egyptian and Standard dialects.
• Addressed challenges like dialect similarities (Saudi/Khaliji) and data imbalance through stratified sampling and hyperparameter tuning (Optuna).
• Deployed the model via a FastAPI backend and a user-friendly web interface for real-time dialect classification.

Pneumonia Detection from Chest X-Rays 

• Developed a CNN-based deep learning model to detect pneumonia from chest X-ray images, achieving 92.5% validation accuracy
• Designed a preprocessing pipeline using OpenCV for image normalization, contrast enhancement, and resizing
• Applied class balancing and data augmentation techniques to reduce false negatives by 18% compared to baseline performance
• Evaluated model performance using precision, recall, F1-score, and confusion matrix for medical reliability

Impact of Boycott Movements on Starbucks Stock Price

• Used historical stock data from Kaggle to analyze the effect of boycott movements on Starbucks' stock performance
• Applied time-series modeling to assess both short-term and potential long-term price trends resulting from public boycotts
• Observed a noticeable drop in stock price during boycott periods, with projections indicating further decline if boycotting continues
• Analyzed online sentiment to support market trend observations and consumer behavior patterns
• Presented findings through data visualizations and reports highlighting the financial influence of collective consumer action

Exploratory Data Analysis for Machine Learning