Amazon Bedrock Data Automation launches instruction optimization for your document blueprints

Amazon Bedrock Data Automation (BDA) now supports blueprint instruction optimization, enabling you to improve the accuracy of your custom field extraction using just a few example document assets with ground truth labels. BDA automates the generation of insights from unstructured multimodal content such as documents, images, audio, and videos for your GenAI-powered applications. Blueprint instruction optimization automatically refines the natural language instructions in your blueprints, helping you achieve production-ready accuracy in minutes without model training or fine-tuning.

With blueprint instruction optimization, you can now bring up to 10 representative document assets from your production workload and provide the correct, expected values for each field. Blueprint instruction optimization analyzes the differences between your expected results and the Data Automation inference results, and then refines the natural language instructions to improve extraction accuracy across your examples. For your intelligent document processing applications, you can now improve the accuracy of extracting insights such as invoice line items, contract terms, tax form fields, or medical billing codes. After optimization completes, you receive detailed evaluation metrics including exact match rates and F1 scores measured against your ground truth, giving you confidence that your blueprint is ready for production deployment.

Data Automation blueprint instruction optimization for documents is available in all AWS Regions where Amazon Bedrock Data Automation is supported.

To learn more, see the Bedrock Data Automation User Guide and the Amazon Bedrock Pricing page. To get started with blueprint instruction optimization, navigate to your blueprint in the [Amazon Bedrock console](http:// get started with blueprint prompt optimization, navigate to your blueprint in the Amazon Bedrock Data Automation console, go to your custom outputs for documents, and select Start Optimization. ), go to Data Automation, select your custom outputs for documents, and select Start Optimization.