"I 've done it so it's not so bad" is not a very good argument. I've done a significant data science project in Prolog (with R for the plotting btw) but that doesn't mean Prolog is the go-to language for data science :)

Mainly because production pipelines care a great deal about performance (speed) and Python is generally considered to have worse performance in comparison to compiled languages (Java or C++.) Depending on your pipeline that may not be a big deal.