Responsibilities include:

• Training PyTorch Geometric GraphSAGE GNN to classify partnerships’ risk of noncompliance using networks of taxpayer structures
• Implementing self-supervision tasks on heterogeneous graph such as link prediction to improve performance of GNN
• Analyze model predictions to ensure businesses identified provide greatest opportunity for increased revenue
• Utilize current ML and accounting research to guide decision making to improve the IRS’s audit selection process

Responsibilities include:

• Analyzed XGBoost auto underwriting model to identify key areas for improvement, increasing predicted profit from model by 28%
• Created 15 new features using SQL queries, resulting in a 23% increase in predictive accuracy for high-risk policies
• Optimized models using SHAP and XGBoost feature importances, maintaining performance after removing 60 features
• Built R-shiny dashboard with lift charts and profit improvement visualizations, facilitating decision making by stakeholders

Responsibilities include:

• Implemented contract testing framework, allowing for scalable testing of all microservice applications within data pipelines
• Developed 2 Rest APIs and updated functionality of existing APIs responsible for handling $2 trillion in consumer payments daily
• Built a new microservice to maintain logs using Java, Spring Boot, and Kafka

Responsibilities include:

• Automated 6 data pipelines using ETL framework, ingesting and transforming consumer data using Spark SQL and JPMC libraries
• Tested pipeline transformations and implemented step definitions using Cucumber files and deploying to DPL server