AWS Platform Engineering Agents

The phData AWS Platform Engineering Agent is one of the first production-minded platform agents in phData’s broader Intelligence Platform suite, focused specifically on data engineering and modeling. Point it at a transactional PostgreSQL database and the agent will automatically discover tables, columns, primary/foreign keys, and row counts in real time via the phData Forge Toolkit. From there, users can ask business questions in natural language, and the agent generates and runs the SQL on the fly against the schema it just learned; no prior training on that database, no hardcoded joins.

Where it really differentiates is in model design and code generation. The agent analyzes the discovered schema to identify business processes and propose a Kimball-style star schema: fact tables with clearly defined grain, conformed dimensions (products, customers, employees, territories, etc.), and documented key relationships plus the business questions those models support. It then uses this design to generate a complete dbt project (sources, staging, marts, tests, macros) and a MetricFlow-compatible semantic layer, which compiles cleanly and is ready to plug into your existing analytics stack.

All of this runs on AWS-native foundations: Amazon Bedrock AgentCore orchestrating tools, Strands Agents SDK for agent behavior, the phData Toolkit for scans and profiling, and a simple bootstrap script that provisions the AWS environment (VPC, security groups, subnets, RDS instance, seeded Northwind dataset, config, and Python dependencies). The result is a repeatable, infrastructure-as-code–friendly pattern for standing up intelligent, self-configuring data engineering agents that collapse weeks of manual discovery, modeling, and dbt authoring into a guided, interactive experience.