About This Talk
At the dbt Meetup 2024, Leonid Kozhinov and Nikita Yurasov — the creators of the rapidly growing dbt-af integration — discussed working with big data in the data mesh concept. They described their tech stack, the available options for running dbt jobs, and gave an honest assessment of their library’s strengths and weaknesses.
Key Ideas
Data Mesh in Practice — The talk covered how data mesh principles translate into concrete engineering decisions: domain ownership, data as a product, self-serve infrastructure, and federated governance — all implemented with real tools rather than abstract frameworks.
The dbt-af Tech Stack — A deep dive into the technical architecture behind dbt-af, including how dbt models map to Airflow DAGs, how dependencies are resolved across domains, and how the system scales with growing data volumes.
Options for Running dbt Jobs — Not every team has the same infrastructure constraints. The talk explored different execution patterns for dbt — from simple sequential runs to fully parallel domain-based execution — and when each approach makes sense.
Honest Assessment — Rather than just showcasing success stories, the speakers candidly discussed where dbt-af falls short, what edge cases are tricky, and where the integration needs further improvement.
Why It Matters
Open-source tools thrive on honest feedback from practitioners. This talk provides a balanced view of dbt-af — what works well, what needs work, and where the project is headed — helping teams make informed decisions about adopting the library.
