Conducted exploratory data analysis (EDA) on 21,000 Mexican properties using pandas, matplotlib, boxplots, and seaborn, assessing price influences to provide data-driven purchase insights.
Developed ARIMA and ARMA models with hyperparameter tuning in Python (pandas, plotly express, matplotlib, seaborn, statsmodels) to forecast PM2.5 concentrations in Nairobi and Dar es Salaam, supporting proactive pollution control.
Built Logistic Regression and Decision Tree models with scikit-learn from SQLite data to predict building damage from the 2015 Nepal earthquake, identifying low-strength materials as a major risk factor.
Deployed Decision Tree and Gradient Boosting models using Python (pandas, matplotlib, plotly express, seaborn, scikit-learn) to predict company bankruptcy in Poland and Taiwan, aiding strategic investment decisions.
Performed end-to-end customer segmentation using k-means clustering on 20k+ U.S. household records (351 variables), resulting in 4 actionable segments and an interactive Dash dashboard for non-technical stakeholders.
Designed and executed a randomized A/B experiment using MongoDB applicant data, performing chi-square analysis and building ETL pipelines and an interactive web app to inform evidence-based decision-making.
Built and deployed an end-to-end volatility forecasting pipeline for MTN Group equity data, integrating API-based data ingestion (Alpha Vantage), SQLite storage, and GARCH(1,1) modeling, improving reproducibility and enabling programmatic forecasts.