At Wuppertal Institute for Climate, Environment and Energy, I collect, clean, and validate datasets across Nigeria, then analyse them and share findings with stakeholders in Germany and Nigeria. I also contributed major sections to research manuscripts and received a EUR 2,500 international research grant.
I took my Real-Time Fraud Detection System from model development to production, using class-weighted XGBoost to detect fraud in highly imbalanced transaction data. The deployed FastAPI service runs at roughly 15ms latency and includes MLflow tracking, a Docker container, and a monitoring endpoint.
For CreditIQ, I built a loan default prediction system with an ensemble model and SHAP explanations at both global and individual-borrower levels. I deployed it with a Streamlit dashboard and FastAPI endpoint.
My research brings econometrics and machine learning together. In a study forecasting Nigerian GDP across 50 years of data, I compared OLS, Random Forest, XGBoost, and ARDL; XGBoost achieved an RMSE of 0.028 versus 0.224 for OLS.
