About
I'd rather show you where my model is wrong than hide it behind one AUC.
I'm Tjay Burger, a final-year Computer Science and Statistics student at North-West University in Potchefstroom. I build models for decisions under uncertainty: credit scorecards, causal uplift, volatility forecasts, market regimes.
01 · Who I am
Every project on this site has a section where the model loses. The alternative scorecard has a lower Gini than the bureau one, and the page says so. The T-learner ranks better than the causal forest but lets more Sleeping Dogs into its top decile. I think that is the useful part.
I write the core logic myself when it matters. The WOE binning and points scaling in the scorecard engine are from scratch, no scorecardpy. Then I ship it: FastAPI on Render, React on Vercel, Streamlit when speed matters more than polish.
Off the keyboard: third at the South African wrestling championships (Olympic-style), rugby, daily training, and a Hyrox race coming up. Class rep, and group leader on my university projects.
02 · Why hire me
01
Interpretable first
A scorecard a committee can sign comes first. If XGBoost wins (0.753 vs 0.737 AUC on Home Credit), it wins on the same features with SHAP beside it, and the scorecard still makes the decision.
02
Treatments, not just grades
Limit increases and offers are interventions. I use uplift models to find who the treatment moves and who it harms, and I test them on data where the true effect is known.
03
Time series done honestly
Walk-forward splits, quantile forecasts scored with pinball loss and coverage, and a dumb baseline every model has to beat. No random splits on market data.
03 · The book I know
South African data is part of most of this work. Six JSE banks for volatility, JSE stocks next to crypto for regimes, ten SA employers' career sites for GhostCheck, and POPIA rules on what I'm allowed to store.
04 · What I am looking for
Exams end 16 November 2026. After that I'm moving to the Western Cape and looking for junior quantitative ML or credit-risk roles in Stellenbosch, Cape Town or remote.