I never touched FORTRAN after about 1986. Even in the 80s and 90s the science guys complained about fundamental numerical problems with C. Not sure of that still holds or whether Python is better. Most of what I've done is integer and non-demanding floating point numerical stuff.
I have had to use MATLAB before. While it works well for some things, it is horribly inefficient. One job I started out inheriting a mess of a computer vision algorithm that had also been implemented in MATLAB (for a multi-million dollar piece of production capital equipment). What a disaster. After a year I was finally able to convince management that both the algorithm and its implementation language were terrible choices. Redesigned it and implemented in C/C++ for a 10-30x speedup.
For one-off heavy lifting scientific computation/modelling stuff it might be a valid choice, but you will pay in time and/or compute resources if it is a big computation. In production, throughput (and cost) are king.