Modernize and migrate on-premises fraud detection machine learning workflows to Amazon SageMaker

This article details how Radial, a 3PL fulfillment provider, modernized their on-premises fraud detection machine learning workflow by migrating to Amazon SageMaker, achieving significant improvements in efficiency and performance.

• Legacy workflow involved manual, time-consuming model development and deployment processes taking 2-4 weeks
• Migrated to a multi-account MLOps architecture using SageMaker, GitLab, Terraform, and AWS CloudFormation
• Implemented a secure, scalable workflow across development, pre-production, and production environments
• Ensured data privacy and compliance using AWS Direct Connect, VPC, and KMS encryption
• Achieved 75% reduction in ML model deployment cycle and 9% improvement in model performance

The migration enabled Radial to dynamically scale infrastructure, accelerate model deployment, and maintain consistent model training across environments, demonstrating the benefits of modernizing ML workflows on AWS.