If given a modern tooling, Fortran is not too bad, thought. It is in some respect a "no bullshit" language which doesn't allow you to spend your time on infrastructure, architecture or meta programming, because it is virtually impossible to do these things within Fortran ;-)
(I am the original author of LFortran.)
Although it's general-purpose, Fortran skews heavily towards numerical calculations, which is probably why it's not used much outside science and engineering (e.g. most Web stuff is 'string processing' rather than numerical).
Fortran was used to write highly optimised numerical libraries like BLAS, which are so widely used and performance-critical that it pretty much cemented Fortran's use; although (a) such libraries can be used by other languages, and (b) they've become so optimised over the decades that they're essentially pure machine-code these days ;)
(Not that I'm judging, people should use whatever language they prefer).
Despite what you may have heard, the language has evolved over the decades and has most of the features one would expect from a modern language.
Edit to say I missed your newer project specification; MCNP is still under very active development and frequent updates, if that counts.
It even got modules before C++.
I get a bit annoyed by Fortran advocates complaining about people using C++/Python instead of it, when there aren't libraries for doing really basic things. I think Fpm is a really good step in the right direction though.
To be honest reading https://github.com/fortran-lang/fpm/blob/master/src/fpm_back... the only thing look like the old fortran I know is integer I j and do loop. No go to I guess.
NumPy does not use Fortran. It provides an interface to BLAS (Basic Linear Algebra Subprograms) and Lapack (Linear Algebra Package) for which the reference implementation is written in Fortran.