TechPrufer
Sign inGet started

Blogs›Amazon›ML-driven Fraud Detection with Graph Enrichment

ML-driven Fraud Detection with Graph Enrichment

ML-driven Fraud Detection with Graph Enrichment

1
posts
2026

This feature thread tracks the evolution of building and deploying machine learning models for fraud detection, with a specific focus on integrating graph-based features derived from complex relationships within data. Initial efforts focused on foundational ML model development for structured data. This post introduces the modernization of fraud claims processing by combining graph-based features with ML models, leveraging Amazon EMR Serverless for scalable processing, Apache Iceberg for data lakehouse management, and integrating with claims handling systems like Guidewire for actionable insights. The architecture emphasizes a layered lakehouse approach, robust orchestration with Amazon MWAA, and secure integration patterns.

2026

How Mapfre Insurance modernized fraud claims with Amazon EMR Serverless | Amazon Web Services

7/14/2026

This post details Mapfre Insurance's modernization of fraud claims processing by integrating graph-based features with ML models on AWS. It introduces a technical architecture built on Amazon EMR Serverless, Apache Iceberg, AWS Glue Data Catalog, and AWS Lake Formation. The solution utilizes Neo4j for graph enrichment and Apache Airflow on Amazon MWAA for orchestration. A key contribution is the resilient integration with Guidewire Claims via AWS Lambda, including retry mechanisms, dead-letter queues, and secrets management using AWS Secrets Manager for real-time fraud alert actionability.