, Principal Solutions Architect, Amazon Web Services
, Head of Payment Networks, AWS Worldwide Financial Services
Transaction fraud detection is a $43 billion annual problem. AI for fraud detection uses multiple machine learning models to detect anomalies in customer behaviors and connections, as well as patterns of accounts and behaviors that fit fraudulent characteristics.
We'll discuss and share the optimal workflow with data ingestion, data processing and transformation, feature engineering, model training, model inference, and event sequence and notifications running on various open source ML frameworks, NVIDIA AI tools, Triton Inference Server on Amazon EKS or EC2 instances on NVIDIA GPUs. The cost to build features to train AI racks up into hundreds of millions of dollars. NVIDIA RAPIDS and RAPIDS Accelerator for Apache Spark on Amazon EMR enable data processing to be 14X faster at 8X lower cost, compared to CPUs based on fraud detection data benchmarks.