How much math you need to work as MLE in a company but not academia?

How much you really need to know to be successful MLE in a company that focuses on delivering working ML systems and not scientific papers?