Commercial Product Pricing:•Lead a group of Data Scientist and Actuaries in developing rate making GLM/GLMM model for Commercial Farm Ranch Portfolio, reducing loss cost of the targeted Commercial Portfolio by approximate 10 million.Insurance Cross-sell Up-sell Recommender Model•Lead a group of Data Scientist to build product recommender model for cross-sell/upsell, utilizing GCP infrastructure for production model deployment, increasing cross-sell/upsell for both existing and new customers by 3 million yearly.Customer Acquisition:•Lead a cross-functional pod to develop and deploy propensity to bind model for insurance agent to outreaching and quoting insurance product. Utilize AWS cloud infrastructure to deploy and manage the model, increasing quote/bind ratio for agents by 30%.Research:•Conduct research project on territory/geo segmentation for underwriting, and loss forecasting, present report and territory factor to Actuary stakeholder.Leadership/Outreach: * Collaborate closely with UW-Madison professors as part of Amfam Data Science Institute * Mentored rotational associate data scientists

• Led a high-performing team in the development of a cutting-edge digital marketing campaign look-alike model utilizing deep learning techniques such as RNN and LSTM. This model played a pivotal role in boosting ROI by 30% and elevating the conversion rate by an impressive 25%.• Independently engineered generative forecasting models, including Naive Bayes, Bayesian Network, and Markov Chain, integrated with a scalable in-database analytics library (PostgreSQL Madlib). These models successfully reached a population of 350 million in the United States, resulting in a remarkable 17% improvement in conversion rates for Direct Mail and email campaigns.• Conducted sophisticated survival analysis to predict customer time-to-purchase, leveraging methods such as time-varying survival regression and Cox proportional hazard models, with the aid of lifelines and scikit-survival packages. This initiative led to an outstanding 230% enhancement in the 1st decile lift.• Spearheaded the development of a Credit Card ITA (Instantaneous Transaction Authorization) model, meticulously fine-tuning hyperparameters to optimize both performance and complexity. Conducted rigorous A/B testing to assess various marketing strategies and effectively communicated model outputs to business users.

• Led the Model Validation efforts as a subject matter expert for Consumer Loss Forecasting models in the context of CCAR (Comprehensive Capital Analysis and Review). This included Credit Card Probability of Default (PD), Loss Given Default (LGD), and Exposure at Default (EAD) logistic regression models. Key responsibilities involved conducting snapshot back testing, performing sensitivity analysis under various CCAR scenarios, assessing segmentation effectiveness, analyzing variable sensitivities, overseeing model implementation, and establishing continuous monitoring. Regularly presented findings and comprehensive reports to the Model Validation Director and Regulatory authorities.• Supervised and mentored a team of Junior Validators in the validation of Anti-Money Laundering, Fraud, and OFAC sanction screening models. Leveraged R and SAS for meticulous data reconciliation, historical alert regeneration, and sampling alert generation for back testing. Conducted in-depth analysis of missed fraud cases to evaluate the performance of fraud detection methodologies and meticulously documented validation reports.• Conducted rigorous validation of Market Risk and Treasury models, spanning areas such as commercial and residential mortgage servicing rights, Economic Profit Added, and unemployment rate models. This involved a comprehensive assessment of assumption validity, rigorous back testing, sensitivity analysis, and meticulous cash flow calculation replication from QRM (Quantitative Risk Management) for CCAR compliance and reporting purposes.

• Utilized advanced machine learning techniques to create Anti-Money Laundering models based on tree-based algorithms.• Engineered a Fraud scorecard model, implementing variable dimension reduction and model selection strategies. Conducted thorough performance back-testing and established an ongoing monitoring framework.• Executed AML solution Mantas scenarios, specifically focusing on the High-Risk Counterparty Cash and Wire Transaction scenario. This included fine-tuning thresholds through sampling and optimization processes.

•Develop Chinese Natural Language Processing prototype with machine learning techniques using R. Research area include word segmentation, part-of-speech tagging, syntactic and semantic parsing.•Implement OFAC list screening software FinScan for major Banking and Insurance clients using SQL.