How Sigma runs a 2,883-model semantic layer on dbt and Dagster

Sigma Computing operates a 2,883-model semantic layer as version-controlled code in dbt, with Dagster orchestrating builds and deployments across six data regions. Lead data platform engineer Matt Senick explained that the layer is intended to provide AI applications with policy guardrails, consistent metric definitions, and richer context such as synonyms and verified queries, allowing agents to retrieve standardized metrics rather than rely on generic text-to-SQL generation. Semantic objects are written in SQL and Jinja, mapped to warehouse tables, controlled at the column level, and organized by business functions such as sales, marketing, customer success, and product. Sigma uses dbt packages to compile these definitions into Sigma data models, Snowflake semantic views, and Cortex agents, while 63 Dagster jobs manage manifest retrieval, state-aware dbt builds, API synchronization, and manifest uploads. Senick also introduced Open Semantic Interchange, or Ossie, an initiative intended to enable conversion among semantic formats from Snowflake, dbt, and Sigma without requiring a single vendor’s model.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.