R and Python vetiver package: a suitable MLOps solution?
Should you use vetiver as a free, open source MLOps solution?
Today we will be dissecting the vetiver package, available for R and Python. We will look at its core functionalities and we will see how they stack against the traditional MLOps requirements. Armed of this new knowledge, we will see if we should place our bets on vetiver as a suitable MLOps solution for our work environment.
Note that we will mention R in this article, but everything we discuss can be applied to Python.
What is vetiver?
vetiver was first published in CRAN at the end of 2021. It then gained a lot of traction in 2022, following a presentation at the annual RStudio/Posit conference.
R has plenty of advanced tools to retrieve and explore data, and to develop machine learning models. Just think of tidyverse and tidymodels. What was lacking was an elegant way to document, to deploy and to monitor the models. Enter vetiver.
In vetiver’s official documentation we can read:
The goal of vetiver is to provide fluent tooling to version, share, deploy, and monitor a trained model.
To the trained ears, this sounds a lot like “MLOps”.
