Prompt to Pipeline: Hands-On Data Engineering with Cortex Code

Your AI Wrote the SQL. Did the Data Land?

A 2-hour hands-on lab: build a Data Engineering Assistant with Cortex Code

Data engineering work rhymes. Every day brings the same pipeline debugging, the same migration grunt work, the same environment checks. Never identical, never truly new.

Generic AI coding agents can write the SQL. They can’t tell you whether the data arrived, what your change will break downstream, or why your warehouse is spilling. They see your code. They don’t see your data.

So you re-explain your environment every session and check every result by hand.

Cortex Code (CoCo) is different. It’s connected to your Snowflake account, so it doesn’t stop at writing code. It runs it, checks that the data landed, reads lineage to see what a change will break, and pulls metering to show what it costs.

In this lab, you’ll turn CoCo into your team’s own data engineering assistant: one that knows your conventions, remembers your environment, and stops asking the same questions.

What you’ll build

• An agent that knows your environment. Capture your team’s conventions once in an AGENTS.md file, so you never re-explain your setup.
• A morning-check skill. Turn your daily pipeline review into one command that shows what’s broken, stale, or costing you money.
• An end-to-end pipeline, built and verified. Use CoCo to ingest with Openflow, transform with Dynamic Tables, and serve analytics through a semantic view.
• A plan for trusting it with more. A simple framework to go from diagnose to verify to operate, so the agent takes the routine work and you keep the decisions that need a human.

You’ll leave with a working assistant you can use Monday morning.

This lab is for you if:

• You re-explain your schemas, conventions, and environment every single session
• Your mornings start with «what broke overnight?»
• You own pipelines in Snowflake and want the agent to do the grunt work, not just suggest it