You should definitely learn Python. Even Python is often considered as a prototyping language. Many production models are converted from Python to languages like Java or C++. I don't think that neural networks can be effectively coded using JavaScript. However, this is possible because the neural network is nothing that the mathematical model and any programming language can be used to perform mathematical computations.

I wouldn't be so confident about the impossibility to use JS for this. Google recently released Tensorflow.js. This library can be used to perform neural networks training on JavaScript. Moreover, it can be executed on the client-side, directly in the browser! This is amazing. I think we are moving towards developing ML solutions using convenient scripting languages that can operate on the frontend rather than on the backend.

But JavaScript cannot provide such speed that can provide Python!

Python is also not very fast. Java and C++ are faster. Nevertheless, Python is the most popular language for Data Science. Computers become more and more powerful, and I think that the speed of the specific programming languages matters less now than even five years ago.

I agree with (B) that JavaScript is not the standard language for deep learning. Nevertheless, it is possible to do this, and if you don't want to learn Python, you can try to use JS. The particular use case is also matters. How many training data you have, what should be the latency, is this the computer vision task or something else, etc. All these things are important when deciding about is it ok to use JS. But I also want to add that there are several other Javascript libraries for neural network creation (except well-known Tensorflow.js). For example, there are BrainJS (https://brain.js.org/), Synaptic (https://github.com/cazala/synaptic), Neataptic (https://wagenaartje.github.io/neataptic/). I'm sure that there are even more, try to find them in Google if you want.

I will try TensorflowJS for the POC. Maybe later, I will move to Python. Thanks, everybody!