Orchestrating data & AI pipelines with Airflow and Snowflake

Building an AI agent is the easy part. Running it reliably in production is harder: the agent needs trusted, up-to-date data, and the pipeline around it needs retries, monitoring, and a person who can check the output before it goes further.

Snowflake Cortex Agents provide the reasoning, working over your structured and unstructured data inside Snowflake's governed environment. Apache Airflow® provides the production orchestration. The Airflow Snowflake provider lets you create, run, and manage Cortex Agents as Airflow tasks, right next to the SQL that loads and prepares their data.

In this webinar co-hosted with Snowflake, you'll learn how to run end-to-end data pipelines that load data with Snowflake SQL and run Cortex Agents in production, with retries, observability, and human approval from Astro, Astronomer's managed Airflow platform.

You'll learn how to:

• Keep agent tasks from failing silently with, using Airflow retries and observability that surface the error when a run breaks
• Add a human approval step with Airflow's Human-in-the-Loop operators, so a person signs off on agent output before it moves downstream
• Get end-to-end observability by running Snowflake SQL loading, data quality checks, and agent calls in a single Airflow pipeline
• Ground Cortex Agents in governed data by validating it with data quality checks before agents use it