Conclusion

• Python’s ML(AI) ecosystem is mature.
• Python’s ecosystem provide easy to use ML.
• Other language is getting useful.
• Use managed service at first, choice another lang when not satisfied.
• Take right choice for yourself.

Python’s ML ecosystem is mature.

This is very common sence within these ML(Deep learning) and AI era.

It is offcourse OSS and free to use. It is sufficient for use, even if it is less feature or elegance than commercial math products when look at details.

python’s ecosystem provide easy to use ML.

Pandas, Numpy and so on, it is very useful for exploratory and quickly search without any explain. And that systems based on Jupyter notebook, lile original or Google’s Colab or IDE Supported.

Other language is getting useful

But when integrate Product or online predict api server, it is more good choice rather than Python when consider total cost of performance whole project life cycle.

It is very easy and quickly to start when use web based service like Colab, and make api with flask and cloud service. Some trouble is appears soon.

Python is easy to use and good language but now, we have another choice like,

• Julia
• Rust
• Go
• Haskell

and so on. These are native performance is good and support(or binding is provided) ML, AI.

Use managed service at first, choice another lang when not satisfied.

Google AI Platform is full managed service of ML pipline. There are many service not only google available.

Production level ML is heavy task, so consider to use them at first. Online work(with server) is suitable to call service’s api, we could choice any server tech each likes.

When it is not suitable for requirement, consider to implement yourself. This situation maybe needed tough model design and implement pipline or service level.

Then we could use without Python, rather than more friendly lang for oneself to implement new thought directry and safely.

Take right choice for yourself

Python is open for everyone. And easy to use(battery included) when use prepared service. But the cutting edge that not implemented anyone, there are better choice.

I was same situation, I implemented by myself at first, but trouble was bubbled and system got less robust. And I decide separate ML learn part and online part and I noticed that Python is not always necessary if it is about ML(AI).

Have a good ML(AI) life with your best choice.