A full catalogue of AI, ML, and data engineering work — from NASA space challenges to enterprise-grade prediction engines. Every project built with production quality in mind.
Enterprise-grade Machine Learning solution predicting high-value customer churn before it happens. Built with XGBoost for classification, K-Means clustering for customer segmentation, and SHAP explainability so non-technical stakeholders understand every prediction. The interactive Streamlit dashboard delivers live risk scores, churn probability distributions, and feature importance visualisations — all deployable in a single Docker container.
Advanced weather prediction system developed for the NASA Space Apps Challenge — utilising Machine Learning and massive geospatial satellite datasets to solve critical space exploration and Earth observation challenges. The system processes multi-source atmospheric and satellite imagery data through a TensorFlow deep learning pipeline, achieving high-precision weather forecasts at regional scale. Deployed on AWS with automated data ingestion from public NASA APIs.
Advanced, interactive machine learning diagnostic dashboard with fully segregated Patient & Doctor portals. The Patient portal predicts 48 diseases with 99.5% accuracy from symptom selection, includes an ADA-compliant diabetes agent with offline PDF table parsing and image OCR, a non-invasive Type 2 risk questionnaire, and a bilingual English/Arabic UI. The restricted Doctor portal unlocks live clinic statistics, a patient database with notes/prescription editing, dataset explorer, and searchable diagnostic history logs with one-click Excel export.
Resilient, testable end-to-end web scraping pipeline for Amazon search results and product pages — built with the scrapling framework which uses curl_cffi browser-fingerprint impersonation and randomised headers to bypass Amazon's aggressive bot detection. The parser is fully decoupled from networking (zero network code in parser.py), making it independently unit-tested against saved HTML fixtures. All scraped data is stored via SQLite upsert logic, exported to CSV, and fed into a Pandas analytics engine that generates price histograms, ratings scatter plots, and top-reviewed product charts via Matplotlib.
End-to-end Natural Language Processing system engineered to classify and analyse sentiment across massive datasets of movie reviews. Implements both classical NLP (TF-IDF vectorisation, Naive Bayes, Logistic Regression) and deep learning (LSTM, Transformer-based embeddings) approaches — with a comparative evaluation pipeline so you can see exactly where each technique wins. The system handles text preprocessing, class imbalance, and deploys as a REST API endpoint.