Member-only story
How to Better Deploy Your Machine Learning Model
My argument against homemade solutions and an alternative proposal
Every week, I spend a considerable amount of time learning about machine learning (ML) from both the research side since I'm a Ph.D. student and the engineering side since I've worked as a software developer for quite some time now. While browsing, I keep seeing these tutorials on deploying ML models using some python library that claims it is the "fastest" or the easiest. However, I think that using a python web framework to deploy your ML model is not the best choice out there.
This post will explain why I think you should not rely on a python web framework for your ML model deployment and suggest an alternative. However, this is not a tutorial on using the proposed option. Nonetheless, it is something that I'm working on releasing, so stay tuned.
Why Python Frameworks Are Not the Right Choice
The first downside of using a python framework to deploy your model is that you have to build it. Yes, you can find some repository with a decent template or even a framework on top of a web framework, but you still need to do some modifications. Ask any seasoned software developer, and they will tell you that any new piece of code means the possibility…
