- Led the transition from legacy software to Python, Git, and Sphinx for model development, version control, and documentation. This included implementing best practices for environment management, coding standards (PEP8), testing, and data API design which resulted in dramatically improved transparency, reproducibility, and team collaboration.- Developed Python classes and functions implementing statistical features only available in R or SAS (e.g., generalized variance inflation (R’s car package), truncated power basis expansion (R’s hmisc package).- Initiated centralized knowledge management using “docs as code” approach using Sphinx and reStructuredText. - Reduced deployment time for complex model systems from 6 months to less than 1 week through better tools, design, and collaboration with the Technology teams responsible for deployment.- Founded 400-member Python Users Group and helped grow Group from 6 people to 250 plus in less than 2 years. - Collaborated with Credit, Product, Audit, Accounting Policy, and Capital Adequacy teams to design and develop model systems for Expected Loss (PD/LGD/EAD) for the bank’s Commercial Real Estate and Recreation Finance portfolios for high-visibility regulatory (CCAR/CECL) exercises. Both model systems were successfully validated and passed external audit and regulatory scrutiny.- Directed special projects that explored potential climate change impacts on CRE prices and bank exposure, relationships between homeownership data and CRE vacancies, summarized tax cut implications, and JOBS Act on Bank portfolio. - Presented findings to Executive Leadership at quarterly CRE Emerging Risks meetings.

- Researched new methods and developed proof-of-concept models using cubic regression splines, penalized likelihood (Firth method), the Begg & Grey approximation to multinomial logistic regression, and the Frye-Jacob’s LGD function.- Guided the creation of new models, recalibrated existing models, devised testing simulations, and acted as internal model development advisor. Select successfully completed model systems includes: - PD model for C&I portfolio using competing risks survival analysis. - Recalibrated C&I PD scorecard using logistic regression (“points to double the odds” approach)- Explored new methods and developed proof-of-concept models. Methods include penalized likelihood (Firth method), Frye-Jacob’s LGD function, and using ARIMAX for top-down LGD. - Advised on model development issues related to competing risks survival analysis, stratified sampling, bootstrap and cross-validation, cubic regression splines, p-values, assessment of model discrimination (Somers' D, C-index, ROC, Accuracy Ratio & CAP), and calibration (parametric and non-parametric).- Presented periodic Comprehensive Capital Analysis Review (CCAR) model results to senior leadership.

- Developed the first discrete time competing risks survival model at the bank for the Indirect Auto portfolio. The method was adopted by other consumer and commercial portfolios.- Monitored credit risk of Consumer Installment loans and provided analytical support for underwriting and portfolio management as a Credit Risk Analyst on Consumer Installment team.- Executed vended loss forecasting models for regulatory (CCAR) and quarter loss forecasting. Presented results to business partners and senior management.- Developed training materials for discrete time survival analysis.

Built Linear model to predict changes in cost from schedule delays. Compiled and cleaned dataset for financial reporting using R. Worked with managers to design logic and build a staffing tool.