Fraud and Player Security (Individual Contributor)- Current FocusBuilding ML infra (with a focus on feature store)- Anomaly Detection SystemDesigned, implemented and deployed team's first anomaly detection system that served as a preliminary defense on multiple data platforms which detected and alerted on new fraudulent trends. This system improved overall detection volume on suspicious accounts by up to 3 times, prevented bad actors on further in-game damages.- Data Engineering: Player Profile Feature StoreDesigned and implemented ETL pipelines that aggregate players' snapshot (life-time) behaviors which ease the process for new features generation, model development, operational efficiency, and reduced overall computational time complexities.- Forecasting and PredictionResearched and implemented offline graph (message propagation) based deep learning model and data pipelines that improved the detection of fraudsters by up to 20% while minimizing false positives to less than 1%. In addition, contributed to design document of prevention ML model and feature engineering to harden the criteria for fraudsters to enter EA games in the first place.- SummaryDaily duties include but not limited to: requirements gathering, writing design documents, implementation / development, ad-hoc analysis / modeling, data / feature engineering, mentoring / supporting new team members.

Built POC offline graph based deep learning model for account level fraud detections.

Machine Learning and analytics in operation research

Built Real-Time Unsupervised Machine Learning model pipeline for anomaly detection

Delivered software iteratively (backend focused) using test-driven and agile scrum development methodologies. Played a key role in requirements gathering for user stories, backlog grooming, sprint planning, and development