modeldb enables you to fit machine learning models directly inside databases without pulling data into R. It translates model algorithms into SQL queries using dplyr and dbplyr, working with most database backends.
The package solves memory and performance problems when working with large datasets that live in databases. It currently supports K-means clustering and linear regression, computing everything in the database and returning only the model coefficients or cluster assignments. The model outputs integrate with tidypredict for running predictions in the database as well.
