At Google Cloud Next, dbt Labs is demonstrating how its platform integrates with Google's data stack to address systemic engineering challenges. The focus is on architectural decisions that move beyond basic SQL transformations in BigQuery, AlloyDB, and BigLake.

The core proposition is treating data pipelines as engineered systems. dbt adds version control, automated testing, and active documentation directly into the transformation layer. This approach aims to create a verifiable data foundation, a prerequisite for reliable analytics and machine learning workloads.

A significant technical development is the expansion of dbt's Semantic Layer to work with open table formats like Apache Iceberg. This allows metric definitions to persist across compute engines, reducing the friction when data moves between BigQuery and Spark. The goal is consistent logic, whether the output is a dashboard or a model feature.

For AI-assisted development, dbt is showcasing its Analytics Development Lifecycle (ADLC). This integrates with Google's Antigravity IDE via a VS Code extension, connecting to dbt's Metadata Protocol (MCP) server. The system includes the dbt Catalog for lineage visibility and state-aware orchestration through Fusion to manage costs. The argument is that AI tools need structured metadata and governance to be effective in production.

You can see these integrations at Booth #6606. Demos will cover the Semantic Layer, Catalog, and orchestration. dbt Labs is also co-hosting an afterparty at Eyecandy Lounge on April 22.

Source: dbt Labs Blog