Joshua Harkness

I’m a data science technical leader with 7+ years building risk, response, and customer lifetime-value models for consumer lending—serving credit-constrained borrowers like the population Oportun exists to serve. I’ve built and led full-stack ML functions end to end, owning data pipelines, model training, production deployment, and drift monitoring.

In my current role, I replaced two third-party vendor models with in-house XGBoost risk (charge-off) and response models using bureau data (TransUnion, Equifax, Experian). Those systems influenced ~$240–288M in annual loan originations, with results including ~15% relative reduction in risk and ~5% relative lift in response volume versus incumbent models. I also designed a Customer Lifetime Value (CLTV) framework (8–12 models × 3 bureaus) forecasting 48-month outcomes, reshaping how credit and marketing decisions are made across the portfolio.

I consistently become the technical lead where standards and tooling don’t yet exist—setting engineering practices and raising reliability through model-agnostic automated training, phased champion/challenger rollouts, and monitoring using AUROC/RMSE with thresholds that trigger residual-model refreshes. Previously, I built ML infrastructure modernization on Databricks/PySpark with MLflow pipelines and 10+ Airflow DAGs, and I founded a consumer product business where I owned PostgreSQL, ETL pipelines, and predictive modeling. I bring strong cross-functional partnership and mentorship across marketing, risk, and leadership.