Financial institutions invest a great deal of resources in identifying and preventing financial fraud and money laundering. Due to the ever-evolving complexity of fraudulent activities and the increasing volumes of data, conventional rule-based anti-fraud/money laundering detection systems are overwhelmed. They generate large amounts of alerts, with increasing amounts of false positives and are missing newly emerging schemes due to their technical limitations.

In this webinar, Logical Clocks and NVIDIA will demonstrate how to overcome those limitations by combining regulation-compliant, conventional fraud/anti-money laundering rule-based systems with deep anomaly detection methods--powered by Hopsworks Feature Store, NVIDIA Rapids™, and NVIDIA Triton™ Inference Server. We'll showcase the system’s efficiency in broader low-latency, real-time anomaly detection applications and provide you with step-by-step details on how to develop this state-of-the-art anomaly detection model and manage its entire lifecycle by applying modern MLOps and DataOps principles.

• Develop a state-of-the-art generative adversarial-based anomaly detection model, tune hyperparameters, and train with multiple NVIDIA GPUs in parallel using Maggy on RAPIDS.

• Create an end-to-end anomaly detection pipeline and manage the model's entire lifecycle by leveraging Hopsworks Feature Store.

• Perform model serving with NVIDIA Triton Inference Server and use NVIDIA GPU-accelerated network graphs visualization tools to further investigate suspicious transactions flagged by deep anomaly detection models.

Content

content goes here