Query unstructured data in Amazon SageMaker Catalog using generative AI

This article demonstrates how data consumers can query enriched unstructured data published to Amazon SageMaker Catalog using generative AI, completing a two-part series on governed data discovery and access.

• Consumers search and subscribe to enriched unstructured data assets through SageMaker Catalog's publish-subscribe model
• Option 1: Use Amazon Bedrock chat agent app for no-code natural language queries ideal for analysts
• Option 2: Use Amazon Bedrock model inference for programmatic NLQ integration suited for engineers
• Create knowledge bases from S3 data sources and register S3 locations in consumer projects
• Chat agent returns grounded answers from enriched documents without requiring code
• Jupyter notebook approach enables embedding NLQ capabilities into existing applications

The solution enables organizations to unlock business insights from unstructured data while maintaining governance through SageMaker Catalog's approval workflows.