Build an end-to-end MLOps solution with vetiver for R and Python — Part 1
How can we build an end-to-end MLOps solution with vetiver? And should we do it?
Learning objectives
The goal of this article series is to answer two questions:
Can we create a complete MLOps solution with vetiver?
If so, should we use it?
We’ll answer those questions by creating a toy example. Although everything will be done in R, the same tools and platforms can be used with Python.
Introduction
I previously wrote an article about using vetiver as an MLOps solution. I concluded that vetiver is a viable option only when dealing with a limited number of models and with no strict governance requirements.
While writing that article, I searched for documentation about an end-to-end R-only MLOps set-up based on vetiver. I found a few resources, but they all focus on the model development, rather than on the deployment and monitoring.
So I created my own toy example.
Before we start discussing any technical details, let’s first recap MLOps and the vetiver package.