C++ in DS/ML domains

Technically tensorflow uses c++. Python is used just as a wrapper, it uses c++ behind the scenes.

Either libraries like PyTorch and TensorFlow, or highly optimized code such as at a trading firm. Otherwise there’s very little advantage to coding in C++

ML code written in Python are usually prototypes that need to be further optimized to meet the low-latency requirements in production. So, the answer to your question is yes, but typically only in teams where machine learning is used in production. For example, 1. Tensorflow (as mentioned by other posters) 2. ML models used in Bing is converted from Python to C++. 3. At Tesla, ML prototypes are written in Python or Java and subsequently optimized in C/C++. 4. ML models used in finance are typically converted to C/C++. 5. ML related jobs in VR space also require C/C++.