• Pioneered the development of a global open banking lending product under stringent regulatory constraints and limited data, achieving it within a tight project deadline• Maintained an optimal model accuracy within the 90-95% range through vigilant monitoring, testing and swift adjustments• Established dynamic collaborations with cross-functional teams, including product, software engineering, to seamlessly integrate models and APIs into the broader operational infrastructure• Engineered an 80% reduction in testing time for production updates by developing a custom testing framework, enhancing efficiency• Expanded the data set by pulling and analyzing 40 million transaction records, providing valuable insights for the project team and stakeholders• Mentored 4 junior data scientists on best practices in data analysis, modeling, and communication of findings• Analyzed transaction data to identify distinguished types of credit that impact payment success indicator, resulting in 20% increase in accuracy of credit predictions and improved project outcomes

• Developed a small and medium business revenue stream classification API to help lending process for clients • Generated insight on small and medium business revenue from 100M transaction data using SQL and Python• Upgraded the classification model accuracy by 5% by creating and implementing a business industry identification system• Led 6+ member cross-functional team on small and medium business revenue identification project• Developed a consumer attribute expense stream classification API to create a full view of consumer's finances for client• Improved expense stream classification model identifying black-box problems by performing EDA on 1M+ customer's data• Trained 20+ new hired data scientists on Mastercard data science projects and increased productivity in the team• Managed client dispute questions on API output in a timely manner resulting in higher client satisfaction

• Developed verification of income API based on analysis on 200K+ customer’s transaction data• Created ground truth of income stream dataset by utilizing clustering model and reduced development time by 20%• Implemented z-score based outlier detection method to split outliers in transactions and increased the model accuracy on the outliers by 99%• Built 4 different classification models (XGBoost, random forest, decision tree, logistic model) to create a voting system for income stream classification• Modified logging in income stream API for effective error detection and performance evaluation• Analyzed impact and effectiveness of API update; identified the API speed and overall performance• Reduced projected time for API development by 1 month by cross-functionally coordinating with developers• Designed and tested the application and proposed technology solutions to improve process efficiency