Selected work focusing on applied, end-to-end projects where modeling decisions matter.
Detecting fraudulent credit card transactions in highly imbalanced real-world data where fraud represents less than 1% of transactions. Designed a cost-sensitive evaluation framework that weighted false negatives more heavily, leading to better production-ready model selection.
Anomaly Detection · imbalanced classification · cost-sensitive evaluation · precision-focused
Precision: 0.95, Recall: 0.87
GitHub →Forecasting solar energy production accurately to support grid management and energy trading decisions, accounting for weather variability and seasonal patterns. Built a quantile regression model that outputs prediction intervals at multiple confidence levels, separating deterministic trends from weather-driven variability.
Time Series · quantile regression · uncertainty quantification · ensemble methods
MAPE: 8.5% with well-calibrated prediction intervals
GitHub →Complete machine learning workflow for predicting customer churn in telecommunications, from data cleaning to FastAPI deployment. Includes Monte Carlo simulations for business scenarios and automated hyperparameter optimization.
Classification · end-to-end pipeline · FastAPI · simulation engine
Recall: 0.711, F1-Score: 0.624, ROC AUC: 0.842
GitHub →