Mostafa Ali

I've built end-to-end machine learning pipelines for credit-risk classification and predictive maintenance, translating complex datasets into measurable predictive outcomes.

In my credit-risk project, I developed a modular Python and XGBoost pipeline for loan-default prediction, achieving 93% accuracy and 95% ROC-AUC on held-out test data. I separated preprocessing, feature engineering, training, evaluation, and inference to support reproducible deployment.

Through internships with NTI, DEPI, and Global Course Innovation at Matsuo Lab, The University of Tokyo, I've worked across data wrangling, visualization, SQL, model selection, deep learning, MLOps, and prompt engineering. I've used MLflow to track experiments and model versions while applying supervised and unsupervised learning to business contexts.

I'm pursuing a BSc in Data Science and enjoy rigorous problem solving with Python, SQL, Pandas, NumPy, Scikit-learn, and XGBoost. I focus on careful feature engineering, performance evaluation, and building practical data workflows.