Originally posted on LinkedIn.
During work on my main project I needed a simplistic but performant data warehouse. And as a frugal person I decided not to go with enterprise-grade solutions, but to go with open source.
I chose this setup: Ducklake, SQLMesh, and Jupyter.
I’m genuinely impressed by this combination.
What it gave me
- Fast queries — DuckDB’s performance and simplicity still amazes me
- Clean data modeling — SQLMesh brought solid workflow management to the table
- CI/CD-like data models — version-controlled, testable, deployable
- Intuitive development experience — Jupyter delivered what it should: a classy and polished notebook experience
- Google-scale potential — combined with the Ducklake standard, DuckDB can handle serious amounts of data
Why it matters
This feels like it could be valuable to the broader data community. It addresses some real pain points I’ve had in previous projects — the kind where you need real analytical power but don’t want to spin up (and pay for) a full enterprise data warehouse.
The stack is entirely open source. No vendor lock-in. No surprise invoices. Just fast SQL on your own terms.
Check it out: duckpond on GitHub