Your warehouse already has the data, the lineage, and the tests. So why is your batch ML pipeline living somewhere else?

In this session, Matt Senick (Senior Analytics Engineer, Sigma Computing) walks through a standardized pattern for running the full batch MLOps lifecycle inside dbt: feature engineering, training, inference, model evaluation, and versioned model comparison. No separate Python scheduler. No duplicated feature logic. No broken lineage.

Matt implemented this pattern in production on Snowflake at both 7shifts and Sigma, distilled into three open-source dbt packages - dbt-ml-inline-preprocessing, dbt-ml-eval, and dbt-ml-ab-testing.

You'll leave with a reusable project structure, a decision framework for when this approach fits, and hard-won lessons from two real deployments.

For analytics and data engineers who want to own batch ML workflows without standing up a separate MLOps platform.