How a Health Media Co Built 140 Data Models in 6 Weeks

A 51–250-person healthcare and media company scrapped an original automation request once the real problem emerged (no coherent data architecture) and built a modern data stack from scratch. A CI/CD process built around Claude Code, which wrote roughly 99% of the code and documentation, let the team stack prior session context into the repository so learnings compounded. This produced nearly 140 independent data models in six weeks, work that would have taken six to nine months with a traditional team.

The build used Claude Code to write approximately 99% of the code and documentation, on a stack of Snowflake (warehouse), dbt (transformation), Fivetran (ingestion), the Wallabi platform (application layer), and Tableau (reporting), with Snowflake Intelligence for natural language queries. The approach combined AI-accelerated custom software with data synthesis and reporting.

The team built 140 data models in six weeks on a production-ready medallion architecture, with Claude Code writing 99% of the code. Reporting that previously took weeks became near-instant, and the work became a direct input to the company's board narrative and 2026 growth strategy.

The 140 data models were built in six weeks, within a 4–8 week engagement range, versus the six to nine months a traditional team would have required.

Mid-market companies in services, media, or events that know they need to do something with AI but keep hitting dead ends, especially those that have already bought the right tools and still aren't seeing results.