Hello, world
Dabbler, enthusiast, tinkerer. I build small tools, chase odd questions, and try to uncover patterns hiding in messy data.
Selected work
Classification
Roughly one in four fee obligations at my school went unpaid, and we only found out after the term had already gone wrong. This predicts who is at risk in week one, so the conversation can happen early. Presented at IndabaX Ghana.
Machine learning
Mobile money fraud detection under severe class imbalance, with the full MLOps stack behind it. The interesting part was not the model. It was learning that a beautiful ROC-AUC can hide a useless system.
Product
A CV and cover letter engine built on a knowledge base of everything I have ever done. It finds the transferable parts of my background when a role looks nothing like my last one, and scores how well I actually match.
Currently
MSc in progress, learning in the open
Focus
Classification, forecasting, and honest evaluation
Stack
Python, SQL, scikit-learn, Docker
