Yiwei Cai Email & Phone Number

Kirkland, WA, US

Redmond, Washington, US

Azure Identity.

Montreal, Quebec, CA

• Machine learning fraud prevention system design and development.
• Developed FraudNet, an xgboost-based fraud prevention machine learning model with over 0.985 AUC, reducing review rate by 70% and customer frictions by 65% compared to the legacy rule-based system.
• Production implementation of machine learning model as an REST API on AWS using Docker, uWSGI, Nginx, Flask and its integration with in-house risk platform.
• Created a web portal that provides visual explanation to analysts of model’s decisioning using LIME (Local Interpretable Model-Agnostic Explanations) and seaborn.
• Created scripts to parse, wrangle and clean raw production and 3rd party service JSON events into MySQL using Python SQLAlchemy for analytics and model building.
• Automated daily reports for rules assessment, anomaly detection, analyst performance, void notifications and chargeback reviews through python scripts.

Mentor, Ohio, US

• Built the machine learning model, which can optimize the conversion rate of transactions, recover over 80% of stuck transactions and increase sales by 5% for merchants ($50 million potential increase for Newegg), using.
• Improved classification of bank authentication protocols by 20% by applying clustering methods on transaction data using Python Scikit-learn.
• Modeled risk-based banks’ history behavior of processing time on transactions to automatically create labeled challenge behaviors data using kernel density estimation.
• Reduced customer friction during shopping by 50% and increased liability shift capability up to 25% for merchants by predicting whether the transaction will be challenged before completed using logistic regression and.
• Performed ad-hoc analytic on millions of diverse data, including user agent, latency, geography, devices, and service providers, using Python and SQL to find out potential areas of optimization.

• Researched more than 10 distinct industries, including biopharmaceutical, wearable devices, O2O E-commerce and clean technology, with an emphasis on TMT field.
• Screened over 400 companies and teased out potential target companies for further evaluation using SWOT analysis and Porter five force model.
• Conducted fundamental analysis on target companies, market trend, competitor strength and the bargaining power of consumers and suppliers.
• Joined weekly meetings with investment associates to evaluate potential target companies.
• Assisted investment associates with drafting Due Diligence and deal memo for target companies.

Master Of Science (M.S.), Operations Research

Bachelor Of Arts (B.A.), Finance