Lead Decision Science Team to provide modeling and MIS report supports for personal loan products:  Designed/developed all risk models and pricing strategies for initializing personal loan launching. Prepared all documents and conducted all responses to KPMG validations and investor Elligton validations. The works played an critical role in getting the agreements between loanDepot and investors for the products  Designed/developed daily MIS reports for the personal loans, including risk reports, daily activity report, daily application reports, model monitoring reports, model implementation validation reports, and direct mail and email marketing reports, etc.  Designed/developed loan application waterfall data marts. The data mart provides all data for real-time tableau reports and SSRS reports

J.P. Morgan Chase- Auto Loan Origination and CCAR, Lewisville, TX  Built the loan-level risk-based pricing models using fractional modeling techniques for Chase auto loan C2A project. The Project merges the underwriting engines of prime marketing and subprime-marketing together.  Built PD models using hazard techniques for loan-level CCAR projects. The models predict both loss and timing without simulation.  Provided all analysis supports for the Fair Lending team in the adverse action code development for the new generation origination models (Gen5B) in C2A projects

 Led a team of three modelers to develop state-of-the-art consumer credit/fraud models to improve the bottom line.  Developed direct mail marketing acquisition models • Two dimensional mail selection strategies for targeting high response and low risk consumers. Line-assignment optimizations to maximize revenue given the number of mailers and total principles. • More than 30 million mailers from 01/2012 to 06/2013. Built fraud knock-out rules using downstream anomaly searching to complement upstream models for comprehensive decisions Designed and developed a scorecard toolbox with SAS Macros • An automated system containing a full set of SAS macros for scorecard development. Examples, variable pre-selection, variable stability detection, automatic missing imputation /floor/capping, dimension reduction, manual variable selection supporting tools, and swap set analysis, etc. • The toolbox supports the fast development for all risk/fraud/marketing scorecards with very high quality.

The center assists USDA in combating fraud, waste and abuse in the federal crop insurance system through data mining and statistical analysis. My work included risk modeling (both loss frequency and loss-severity modeling) , risk scoring, fraud detection, as well as, development of new statistical algorithms for spot-check-list development according to data, crop insurance provisions and fraud detection requirements

My work included application of statistics and numerical analysis to environmental sciences, such as environmental risk analysis, sampling and monitoring strategy development, contaminant removal efficiency assessment and contaminant transport numerical modeling, etc.