Senior Quantitative Risk Analyst
Current• Built a charge off prediction model using Markov Chains allowing the business to get a twelve and twenty-four month prediction on demand. • After inheriting the credit-card product, an xgBoost model was created to predict charge-offs - creating efficiency by knocking down prediction run time from eight hours to twenty minutes• Created Monte Carlo simulations for models I personally created to allow business partners to have additional information on the predictions they were being provided• Perform model management / maintenance tasks for three different vendor models while also creating / maintaining my own model creations• Found a bug in a production process that was limiting the number of machine-learned scores we were receiving from a vendor – created a comprehensive report + worked with business partners to go from ~38,000 to ~68,000 scores available + a better understanding on about $2.5 billion in exposure