Exploratory data analysis isn't killer in Haskell. I think types can help a lot here (god knows the amount of pain that R/Numpy spaghetti has inflicted) but the jury is still out for how to do it well. Row types seem like an important technology which gets a little bolted on to Haskell---both exciting and unclear at the moment.
On the other hand, if you've got a solid data pipeline you're scaling then there are grand tools for streaming data in Haskell. You're in the complete sweet spot for strong typing. You can also drop down into C easily when speed is an issue. Matrix libraries exist for the intrepid user. HMatrix is BSD now so you can at least touch LAPACK.
https://acowley.github.io/Frames/
On the other hand, if you've got a solid data pipeline you're scaling then there are grand tools for streaming data in Haskell. You're in the complete sweet spot for strong typing. You can also drop down into C easily when speed is an issue. Matrix libraries exist for the intrepid user. HMatrix is BSD now so you can at least touch LAPACK.