Maybe move away from Python altogether?
At some point, move on, Python started as a toy, and grew into a behemoth.
For some in the scientific community, Julia is seen as a a nice alternative to python. It has some great libraries, and installing them is easy, perhaps because the system was well-designed from the start, and recently enough to build on experience with other systems.
Julia is really handy for reproducible work, because you can specify which versions of which libraries you're using, and share that information with others easily so they can retrace your steps. That's a big factor in scientific work.
The other thing (returning to the main topic) is that Julia does not have to rely on fortran or C/C++ for low-level work. Julia is very fast all by itself. Indeed, most of the Julia libraries are written in Julia itself, providing users with a tutorial for intensive work.
In addition to these things, I have to admit that I love how Julia lets me program in unicode symbols, so my equations in code can look like my equations on paper.
There are some downsides to Julia that I ought to point out.
1. The error messages can be very cryptic. 2. The documentation is poor (no worse than Python, I think, but a lot worse than R). 3. The development community is a bit ivory-tower-ish. The usual answer to "could you improve the documentation" is "please submit a PR". 4. There is a learning curve to get used to Julia. It's easier if you are already familiar with Fortran and Matlab. Then again, most scientists I know are familiar with both those languages (and Python, and R, and ...) so learning a new language is not a challenge. Scientists like to learn stuff -- that's the whole point of being in science, for most of us.