I don't really want to do everything in one language but if I did have to pick it would probably be Julia. It's slightly simpler than Python, and significantly faster without relying on APIs written in C. And has some really great features like broadcasting, multiple dispatch, and a good type system. The only place I feel like Python has it beat is quantity of libraries and support network, which both basically come from the same origin of just having more users. I'm hoping more data science types switch over in the next few years, since Julia is already great for most things mathematical. And I hope that momentum allows Julia to perhaps reach out to other domains.