ORMs and DBT

SQL is King. Long live SQL

Object Relational Mapper (ORM)

ORMs, like Python's SQLAlchemy, abstract SQL-based operations into object oriented programming (OOP) APIs. This achieves a couple of things

1. Devs can now interact with SQL-based data stores without needing to know SQL (mostly)
2. SQL makes it VERY easy to do dumb stuff. ORMs minimize this risk
3. Database migration tools like Alembic are able to intelligently infer the difference between a current database configuration, the new configuration you've described in the ORM, and automatically generate a nearly perfect migration script to "upgrade" the database.
   • Each migration can then be versioned just like software. If something goes wrong, an organization can quickly and safely "downgrade" to a previous version with a single command.

Data Build Tool (DBT)

DBT is far and away becoming a key pillar of modern data engineering. Combine the abstraction of ORMs with the automation provided by Celery, Dask, etc. for SQL-compliant data stores and that's generally where DBT fits in.

It's important to note that DBT by itself is popular but it's really the extensive ecosystem being built on top of or integrating with DBT that is powerful. This is somewhat similar to how Python doesn't necessarily have anything to do with AI... but the ecosystem of tools enabling AI has been built with Python.