How Sigma manages the semantic layer with dbt and Dagster
Sigma Computing's Matt Senick explains how Sigma orchestrates a 2,883-model dbt project with Dagster and manages its semantic layer as code.
Overview
Sigma's data platform team orchestrates all semantic assets (Snowflake semantic views, Cortex Agents, Cortex Search Services, Sigma Data Models) within one dbt project, with a Dagster job that deploys them on change. Previews an open-source Sigma converter for the OSI/Ossie project.
What you'll take away
- 01
Sigma runs its own semantic layer, the same kind it sells customers, as versioned code inside a dbt project: semantic objects defined in SQL and Jinja, compiled into Sigma data models, Snowflake semantic views, and Cortex agents.
- 02
Dagster orchestrates 63 jobs continuously building and deploying across 2,883 dbt models and 6 data regions, syncing results back through Sigma's API on every run.
- 03
The goal is one number every AI surface agrees on: once a question maps cleanly to a defined metric, an agent should call a function and get back the same answer every other surface already trusts.
Session speaker
Matt Senick
Sigma Computing