From Raw Schema to Queryable Data Product: Building an AI Data Engineering Agent on AWS Bedrock AgentCore

See how to build an AI data engineering agent on AWS Bedrock AgentCore. This talk demonstrates automating schema discovery, dbt model generation, and natural language querying of data products, live on AWS.

Overview

AWS Platform Agent is an AI-powered data engineering agent that automates the journey from a raw database schema to a working, queryable data product on AWS — discovering sources, profiling data, generating dbt models and a semantic layer, and answering natural-language questions over the result.

The demo walks through a live Northwinds Financial workload: an 14-schema PostgreSQL source on RDS gets scanned, profiled, and conformed into a dbt staging layer, then deployed behind an Amazon Bedrock AgentCore Runtime with a React + Cognito front end on Amplify. We’ll show the agent reasoning through schema discovery, generating dbt + MetricFlow YAML, and serving a “Talk To Your Data” chat that streams responses through AgentCore Gateway-backed Lambda tools. This follows the AgentCore FAST Template.

Tech stack