I wonder if this type of collaboration is easy to achieve in other languages? Of course not many have an awesome ODE ecosystem like Julia's. But it feels like an unique strengths of Julia where it's easy to integrate heavy duty libraries and have them work together!

It reminds me of this post http://www.stochasticlifestyle.com/why-numba-and-cython-are-...

I think Numba and the Cythan ecosystem will struggle to achieve this with this level of (relative) ease, although I see the original implementation was done in PyTorch. The ODE and ML experts probably made it look easy!

Now I wonder about the relative performance of DiffEqFlux vs the original implementation