Yes, Julia has wrappers for many other languages, so there are options if you really need some specific chunk of C/C++, Fortran, Python, Octave or R etc.

Yet a polyglot project is bad design, and tends to become an abomination in time. This is one reason many Go programmers rewrote libraries rather than saturate their code with cgo/C and SWIG/C++ wrappers. Similarly, people are writing scalable versions of libraries in Julia to improve transparent parallel performance for problems too big for a single machine to feasibly handle.

Things are still undergoing change, but of course the commercial nvidia binary blobs will haunt everyone for awhile. =)