Wei Wang

Experience

• 

  Penn State University

  Aug 2009 - May 2013

  Instructor

  • Introduction to Probability Theory • Introduction to Mathematical Statistics • Intermediate Applied Statistics & Stochastic Modeling

• 

  Fifth Third Bank

  Apr 2013 - Aug 2014

  Quantitative Analyst

  • Developed Bank's credit card portfolio PD and auto portfolio LGD models under Basel II capital adequacy rules for Dodd-Frank Act Stress Testing (DFAST) and Comprehensive Capital Analysis and Review (CCAR) from scratch, as well as to enhance estimations for Consumer credit risk across different business lines. Credit Card PD is an account-level model using Weight of Evidence (WoE) binning transformation, logistic regression and survival analysis; Auto LGD is an account-level model using weighted logistic regression.• Led the development of portfolio-level balance forecast models for non-performing assets (NPAs) using time series analysis. Show less

• 

  Bank of America

  Aug 2014 - Sept 2016

  Quantitative Finance Analyst, Credit Loss Forecasting Analytics

  • Developed Real Estate Owned (REO) NPV model to predict the net recovered proceeds of a normal foreclosure process for a given loan and property, discounted to the present time, which aggregated from component-models including Foreclosure Time, Liquidation Time, Sale price, Attorney and BOP cost, Escrow Cost, Maintenance Cost, as well as Sale Cost. The REO NPV is used either for determining either to proceed or to reject a REO alternative: Traditional short sale negotiation [ensuring short sales mitigate loss to bank/investor];  Foreclosure bidding [improving the proportion of loans going to 3rd party sales]; Stop Advance decision [the judgment of when to cease advancing principal and interest to investors]; Charge off decisions.• Designed and developed Customer Complaint Text Analytics Tool to identify potential noncompliance and to predict customer attrition and balance diminishment after filing complaint. The tool is being used to enhance costumer experience and to increase revenue of the bank:• Used Python to create structure data like unigram, bigram, or trigram variables from a collection of documents contain verbatim; • Applied the structure data on Gradient Boosting SAS macro to analyze the data. Show less

• 

  Wells Fargo

  Sept 2016 - now

  • Responsible for recognizing opportunities to improve the customer experience through the use of machine learning and Artificial Intelligence modelling. Lead data exploration and model development to evaluate and improve the customer experience using unstructured data, structured operational data and customer interaction data.• Manage a global team of data analysts, data scientists, and data engineers in the development of quantitative models and analytic solutions leveraging unstructured data (voice/transcripts/emails/survey, etc.). Show less • Lead independent research project for Natural Language Processing (NLP) including topic modelling of taxonomy, sufficient sample size for machine learning models and interpretability of machine learning models.• Lead the development of Consumer and Small Business Banking complaint identification model and categorization model using phone transcripts to help Wells Fargo to close consent order as well as satisfy expectation from both Line of Business (LOB) and Model Validation.• Design a sensitivity analysis framework to continue improve business rules model performance across all LOBs for both complaint identification model and categorization model. Show less • Developing Pre-Provision Net Revenue Forecasting (PPNR) Model for Dealer Service (Auto) Indirect portfolio using time series methods such as error correction model, including estimation, integration and back-testing of the entire model for CCAR scenarios. The PPNR model is composed of four modules: Credit Models (existing book and new originations), Direct Non-Interest Income and Direct Non-Interest Expense. • Developing loss forecasting models for Dealer Service (Auto) under Consumer Lending Group (CLG) focusing on PD/LGD/EAD models; PD is an account-level model using Weight of Evidence(WoE) binning transformation and logistic regression; Auto LGD is an account-level model using weighted logistic regression.• Documenting models' development processes including methodologies employed and analytics results• Collaborating with business partners to ensure successful development, validation, and implementation of the models• Responding to ongoing analytical requests from auditors and regulatory reviewers Show less

  • Head of AI Exploration and Model Enhancement, Enterprise Customer Excellence

    Jun 2023 - now

  • Quantitative Solutions Engineering Senior Manager, Enterprise Customer Excellence

    Jun 2022 - Jun 2023

  • Quantitative Analytics Manager, Enterprise Customer Excellence

    Jun 2021 - Jun 2022

  • Lead Quantitative Analytics Consultant, Enterprise Complaints Management

    Sept 2019 - Jun 2021

  • Quantitative Analytics Consultant, Consumer and Small Business Banking Credit Modeling

    Sept 2016 - Sept 2019