Well, the main type in data science is "A big bunch of numbers". So unless you're going to encode array dimensionality with your types (possible, but most libraries don't), the domain is pretty close to untyped.

This is why R is the second most common programming language in data science: it really leans into the "Oooh, you got a big rectangle of numbers with labels? Let me tell you all the things I can do with a big rectangle of numbers!"