AWS Bedrock Enables Custom Metrics for RAG & Model Evaluations

AWS Bedrock Adds Custom Metrics for RAG & Model Evaluations

AWS Bedrock Enables Custom Metrics for RAG & Model Evaluations

April 25, 2025 – Amazon Web Services today announced that Amazon Bedrock Evaluations now supports user-defined metrics for both foundation model and retrieval-augmented generation (RAG) evaluations (Amazon Bedrock RAG and Model Evaluations now support custom metrics - AWS).

With this update, teams can:

  • Craft bespoke scoring rubrics by defining numerical or categorical scales that reflect brand voice or domain-specific quality checks—empowering more precise predictive analytics evaluations.

  • Leverage custom judge prompts powered by LLM-as-a-judge, letting you assess conversational AI responses against your own compliance and policy standards—ideal for securing your legal automation workflows.

  • Integrate with real-time dashboards in Amazon CloudWatch or your BI platform to monitor evaluation trends and drill into anomalies—feeding insights straight into our business intelligence playbook.

  • Optimize agentic AI pipelines, using custom RAG metrics to fine-tune chatbots, virtual agents, and support bots—enhancing your customer service AI performance.

“By giving developers the power to define exactly what ‘good’ looks like, Bedrock Evaluations becomes the single source of truth for model quality across any cloud or on-prem environment,” said Swami Sivasubramanian, VP of Data and AI at AWS.

This feature also aligns with best practices for IT & cloud management, allowing DevOps teams to set Service Level Objectives (SLOs) and monitor model health alongside infrastructure metrics—see our IT & Cloud Management guide.

For a full enterprise AI strategy, explore our end-to-end AI Business Automation blueprint.

Enterprises can start creating custom metrics today via the Bedrock console or APIs. For full details, visit the AWS What’s New announcement (Amazon Bedrock RAG and Model Evaluations now support custom metrics - AWS).

Comments

Popular posts from this blog

AI Business Automation: Boost Efficiency & Drive Growth

What Is Artificial Intelligence? Definition, Examples & Use Cases

AI Data Analysis & Predictive Analytics: Tools & Roadmap