Hi @Bangbung

Python is good for fast prototyping. It is convenient. But when you need to deploy models in production you may need to use C++ or Java because these languages are faster than Python.

Regards

But I am working in an organization and our team deploy many solutions written purely on Python. And also I know that many other data scientist use Python for deployment.

There is a project where the speed is not very important. Or it is important but Python is enough. So, Python is actually used for many deployments. But when you are developing something really series and big, it is often needed to use an appropriate programming language (not Python).

Ok. Is there some other reason to use C++ or Java for deployments except the speed of solutions?

Yes. High load. If it is expected that the system will be used by many users in a short period of time then it is better to create this system using Java or C++. Also, the machine learning solution can be a part of the larger ecosystem. Large enterprise solutions are mainly written on Java or C++. So, it is natural to use the same programming languages for ML component in order to avoid artificial connectors between languages. But eventually, it is an architecture choice. Sometimes it is ok to use Python in such cases too.

What about R or JavaScript?

They are similar to Python. But Python is more versatile and convenient than R, that's why it is more popular. I didn't hear about deployments on R. It is a language for research purposes. JavaScript is probably the most popular language nowadays. And it is entering data science too. For example, there is Tensorflow.js for running ML models directly in web browsers. But nevertheless, JavaScript is not as efficient as C++ or Java.