What challenges did BMC face before AI implementation?

Developing a robust fraud detection system to handle high transaction volumes and evolving fraud tactics. Ensuring high accuracy with minimal false positives in payment fraud detection. Building an automated, scalable, and efficient ML pipeline capable of continuous model training, evaluation, and monitoring.

What was the impact of the AI solution?

Improved fraud detection accuracy, with the decision tree classifier showing the most balanced performance metrics. Significant reduction in operational overhead due to automated pipeline orchestration with Control-M. Enhanced real-time fraud detection efficiency through automated, end-to-end workflow management. Facilitated continuous training and evaluation adapting to evolving fraud patterns in financial transactions. Increased ability for financial institutions to detect fraud proactively, reducing financial losses and enhancing customer trust.

Source

Financial Fraud Detection Pipeline with Amazon SageMaker and Control-M

Use case typeFraud detectionUpdated Jun 13, 2026

BMC developed an automated, scalable machine learning pipeline to detect payment fraud with high accuracy and minimal false positives in the financial services sector. The solution uses AWS technologies including Amazon SageMaker, Amazon Redshift, Amazon EKS, Amazon Athena, AWS Lambda, and is orchestrated with Control-M for seamless automation and workflow management. The pipeline automates data extraction from Redshift to S3, preprocessing via Kubernetes on Amazon EKS, model training and evaluation with Amazon SageMaker, and visualization using Athena and Amazon QuickSight. Three ML models (logistic regression, decision tree classifier, multi-layer perceptron) are trained and evaluated to select the best performing one for production. Control-M provides orchestration, scheduling, job dependency management, and automated execution to reduce manual intervention and improve operational efficiency.

Organization: BMC
Industry: Finance
Location: United States
Published: March 2025

Reported outcomes

Strategic outcomes

New product / capabilityBuilt automated fraud detection pipelineCost efficiencyReduced operational overheadSpeed & agilityEnabled real-time fraud detection efficiencyCustomer experience & trustEnhanced customer trust through proactive fraud detection

Primary read

Use case focus

Showing 3 of 3

1Fraud Detection
2Machine Learning Pipeline
3Workflow Automation

Developing a robust fraud detection system to handle high transaction volumes and evolving fraud tactics.
Ensuring high accuracy with minimal false positives in payment fraud detection.
Building an automated, scalable, and efficient ML pipeline capable of continuous model training, evaluation, and monitoring.

Leveraged AWS managed services including Amazon SageMaker for ML training, Amazon Redshift for data warehousing, Amazon EKS for containerized preprocessing, Amazon Athena for querying, and AWS Lambda for data validation and Athena integration.
Used Control-M integration plug-ins to orchestrate the entire workflow from data extraction, validation, preprocessing, model training, evaluation, and visualization pipelines.
Implemented Kubernetes jobs to preprocess raw financial data using containerized Python scripts.
Automated model training and evaluation in SageMaker pipelines for logistic regression, decision tree, and MLP models.
Scheduled monitoring and results visualization through Athena and Amazon QuickSight, enabling real-time insights.

Technologies

Amazon SageMaker Amazon Redshift Amazon EKS Amazon Athena AWS Lambda

Improved fraud detection accuracy, with the decision tree classifier showing the most balanced performance metrics.
Significant reduction in operational overhead due to automated pipeline orchestration with Control-M.
Enhanced real-time fraud detection efficiency through automated, end-to-end workflow management.
Facilitated continuous training and evaluation adapting to evolving fraud patterns in financial transactions.
Increased ability for financial institutions to detect fraud proactively, reducing financial losses and enhancing customer trust.

Architecture

The pipeline orchestrates data extraction from Amazon Redshift to S3, data validation with AWS Lambda, preprocessing on Amazon EKS Kubernetes cluster with containerized Python scripts, followed by automated training and evaluation in Amazon SageMaker. The pipeline outputs evaluation metrics loaded into Amazon Athena and visualized with Amazon QuickSight. Control-M manages the workflow orchestration including scheduling, event triggers, and dependency management.

Implementation partners1

Control-M

Sources & evidence1

Groundedness: 4/5

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence