🐱 How to Build a Local Data Platform with Terraform and Docker?

Author

Grumpy Stack
May 28, 2025

Welcome to my first post!

Hey all!

First off, thanks for subscribing. I publish under the name GrumpyStack — my slightly grumpy personal tech brand focused on all things data engineering. You can also find my blog at p-munhoz.github.io.

I know inbox space is sacred ground these days, so it means a lot that you’ve let me sneak into yours. If you ever have thoughts, feedback, or just want to rant about tech, my inbox is wide open. Seriously. Hit reply — I read everything (unless it’s spammy or weird, then I forward it to /dev/null).

Alright, enough small talk - let's dig into some actual tech, this week: local data platform tutorial, why ChatGPT is making you dumber, and scaling 2000+ dbt models

Table of Contents

This week I published a new piece on my blog.

Think: Airflow, Minio, DuckDB, and LocalStack — all orchestrated like a real grown-up cloud stack, but running entirely on your laptop. Ideal for prototyping, local dev, or pretending you’re a DevOps wizard when you're really just debugging YAML.

If you're curious about:

  • Spinning up a modern data stack locally

  • Using Terraform to manage infra-as-code

  • Learning without burning cloud credits

...then this might just be your new favorite rabbit hole.
👉 Read the full post here

I came across a on how ChatGPT can reinforce confirmation bias — the tendency to seek out info that validates your beliefs (thanks, Wikipedia). This is especially dangerous in tech, where we should be challenging our assumptions, not cuddling with them.

Before reading the article, I already had moments where ChatGPT felt... too agreeable. So I tested it myself:

  • Prompt 1: “I’m the best Data Engineer.” (no extra instructions)

First try without any particular prompt

  • Prompt 2: Same sentence, but with a request for:

    • A neutral, objective take

    • A devil’s advocate critique

    • A positive/encouraging view

Second try with a custom prompt

Yep — big difference. The second one actually gave me a fuller, more helpful response.

Bottom line: ChatGPT is a fantastic tool. But don’t let it become your personal yes-man. Sometimes we need feedback that stings a little.

🔗 How we orchestrate 2000+ DBT models in Apache Airflow?

Because scaling 2,000 dbt models sounds like a good time, right? 😬
This article breaks down how one team makes it work using Apache Airflow.

Highlights:

  1. Mono vs Multi DAG approach

  2. Project structure and DAGs layout

  3. The DAG generation pipeline

  4. How and why we created our DBTOperator

  5. Conclusion and the road ahead

What’s coming next?

I’m working on a tutorial series around Docker for data engineers — but I don’t want to be yelling into the void.

What would you like to read next?

  • A deep dive into something you're stuck on?

  • A beginner-friendly walkthrough of a data tool?

  • A spicy take on tech trends that drive you nuts?

👉 Just hit reply with your idea — even if it’s half-baked or just a single word like “dbt tests” or “Kubernetes pain.”

I read everything, and if your idea makes it into the next issue, you'll get a shout-out (or at least a virtual high-five from GrumpyStack 🐾).

Thanks again for being here — it's genuinely appreciated.

Have ideas or questions? Just hit reply — I might feature your thoughts in the next issue.

Thanks for reading — Pierre (a.k.a. GrumpyStack)
📬 Reply anytime — I might feature you next.