One thing that hasn't been mentioned yet, that I found when I started using Python and SQL together, is the importance of well-considered indices. ORM's aren't always too clever when they generate an index, and often (for repeated queries) you can dramatically cut down the processing time by thoughtfully generating an index.

If I know I'm going to be asking the database a certain question a bunch of times I'll even generate a temporary index to speed up my analysis, and then delete it when I'm done. No sense running a SQL query for an hour if I can cut it down to 5 minutes with an index on the target column.

EDIT: I know this only applies to PostgreSQL, and there are numerous alternatives, but a big shout-out to pghero which helped me identify a bunch of duplicate and missing indexes in one of my databases, saving both time and hard drive space. Incidentally, it also sent me down the road to learn (but definitely not master) index optimization. https://github.com/ankane/pghero