I don't have to worry about the ML API, backend API, and service in python to have correct versions of libraries used, the Python container solution handles that. If it is working on my machine, most probably will work in the deployment environment as well. Review collected by and hosted on G2.com.
Nothing serious, C,C++ level installation of python libraries sometimes fail like NumPy, but this issue is fixable with OS-level configuration fixes and more rigorous testing. Review collected by and hosted on G2.com.