Worth saying that ORMs are very complex, advanced things. They require advanced knowledge of both SQL and Python. If either is lacking, things will be very confusing.

Moreover, unless you are doing some very complicated stuff (eg. large codebase maintained over years), the ORM often creates more busywork than it takes away.

I really would not recommend an ORM as a way of escaping the need to learn SQL.

I second the need to know SQL but ORMs also simplify stuff. It doesn't have to get very complicated to profit from an automatic mapping between objects and records in a database. It takes away the boilerplate code to do this manually or to write an abstraction which ends up being somewhat an ORM itself. Reinventing the wheel.

SQLAlchemy doesn't enforce the ORM though. There is core layer which allows to build SQL queries programmatically without the danger to introduce errors or even security risks, that often comes with handling all with just string operations that contain SQL fragments. Another benefit is the abstraction over differences in SQL syntax of RDBMS. Like which placeholders are used, which characters to quote names, if there is a native boolean type or not, and so on.

Pandas implies SQLAlchemy. The only guaranteed exception that can be used without it is SQLite3.