Real-time user fraud detection

Anomaly detection on large-scale temporal graphs

Developed a heterogeneous temporal graph model for anomaly detection on large-scale time-varying user-attribute graphs with significant class imbalance.

Anomaly detection on large-scale graphs

Developed a large-scale graph-based proactive multi-account abuse prevention system for Buyer Abuse Fixed team at Amazon. Worked on graph construction from large-scale tabular user-attribute data, devising node/edge-sampling strategies for efficient training of graph transformer models to improve recall of anomalous user detection under ~70:1 class imbalance. To give your project a background in the portfolio page, just add the img tag to the front matter like so: