AI Models, Software & Research

Vodafone Czech Republic halved pipeline execution times while using Google’s Data Engineering Agent to convert BigQuery stored procedures into Dataform pipelines.

· ComputeLabs Research · from the November 20, 2026 edition

Google Cloud’s customer account reports that Vodafone Czech Republic reduced pipeline execution times by half while modernizing BigQuery stored procedures with the Data Engineering Agent. The Gemini-powered agent is available in BigQuery pipelines and Dataform; Vodafone used it directly within Dataform, supplying natural-language instructions and code.

The legacy procedures originated in an earlier migration from an on-premises data warehouse to BigQuery. Many contained 2,000 or 3,000 lines of code and ran sequentially, with the case study contrasting that structure against smaller, modular .sqlx files, reusable functions, and Dataform’s `ref()` mechanism.

Vodafone reported that the agent separated code into independent steps that could run in parallel. It also identified table dependencies and cached selected intermediate results using BigQuery materialized views, which the account describes as speeding execution and reducing costs without providing a cost-reduction percentage in the supplied text.

The modular approach supports built-in uniqueness, non-null, and referential-integrity testing, alongside generated documentation and lineage graphs. It also replaces manual dependency management with lineage-based execution through a directed acyclic graph (DAG); the source says Vodafone migrated some stored procedures, not its entire pipeline estate.

  • Vodafone Czech Republic
  • Google’s Data Engineering Agent
  • BigQuery
  • Dataform

All 19 stories from November 20, 2026