Yaoyao Zhang

Master's

Financial Engineering

Ernst & Young

senior analyst

• Engagement: Production Code Development • • Collaborated with IT team in Agile framework to develop a Python library containing 22 modules, covering functions to do data cleaning, conduct complicated data processing, perform model calibration, and produce time series forecast. • • Investigated and debugged a large amount of untested legacy code over 35k lines, reported over 20 code issues in 3 weeks, and integrated legacy code with newly created modules. • • Worked with multiple teams on the development and onboarding of a new automated data quality control program, with the capability of data issue detection, automatic remediation, and allowance for manual overwriting. Helped the bank save the need to recruiting new employees to do manual data checking and improved efficiency. • Engagement: Prototype Development for VaR Model • • Support a global leading Bank in end-to-end delivery of quantitative risk models. • • Used consensus clustering algorithm to support decisions on feature importance ranking and feature aggregation for Bonds attributes. • • Developed tools using Python and SQL with multiprocessing technics to source large-scale market data. • • Developed and implemented new backfilling/infilling models using robust regression and Expectation-Maximization algorithm, which increased data availability and accuracy of historical time series. • • Assessed and improved new models’ performance by using training/validation/testing data sets and choosing optimal set up of models’ hyper-parameters. • • Communicated and interpreted technical concepts to both technical and non-technical client stakeholders, with visualized quantitative analysis results using Python Matplotlib. • Engagement: Model Validation for Retail Checking Account Forecasting Model • • Assessed robustness and performance of OLS and logistic regression models through conducting stationarity test, multicollinearity test, and overfitting test. • • Presented assessment results to the model development team and communicated model limitations.