Why this matters
Most of a data scientist's time goes on preparing data, not modelling. Transactions arrive with text dates, survey rounds need reshaping, and tables need joining. Doing this cleanly in pandas, with checks, is what makes later models trustworthy. Adverts that name Python often name pandas and NumPy alongside it.
NumPy in a sentence
NumPy stores numbers in arrays and applies operations to a whole array at once. pandas is built on it. Whenever you can, write df["amount"] * 1.1 rather than a loop over rows: it is clearer and much faster.
Dates
Convert text to real dates once with pd.to_datetime, then use the .dt accessor: .dt.month, .dt.day_name(), .dt.hour, or .dt.to_period("W") for weeks. Date arithmetic then works: subtracting two dates gives a gap you can turn into days with .dt.days.