MLOps with R: The Whole Game of End-to-End Data Science & Model Deployment

Walk through a machine learning project from start to finish using R, including data preparation, model training, evaluation, deployment, and monitoring. We'll use packages like tidymodels, pins, vetiver, and plumber to take a classification model all the way from exploratory data analysis to a production API. See examples of cross-validation, hyperparameter tuning, creating reproducible pipelines with recipes, publishing model APIs, monitoring model drift, and more. Whether you're new to MLOps or looking to improve your current process, this video will give you the essentials to build reliable, efficient data science workflows in R.

Slides: https://jameshwade.quarto.pub/mlops-in-r-the-whole-game/#/title-slide
Code: https://github.com/jameshwade/r-mlops
Blog Post: https://jameshwade.com/posts/2022-12-27_mlops-the-whole-game.html
Vetiver: https://vetiver.rstudio.com/
Practical MLOps for better models: https://youtu.be/we2hoIyLGXI

00:00 - Intro
00:40 - MLOPs Overview
01:50 - Setup & Packages
03:25 - Exploratory Data Analysis
04:16 - Prepare & Split Data
05:26 - Create a Recipe
07:11 - Specify Models
09:16 - Fit Models & Hyperparameter Tuning
11:09 - Compare Model Results
12:19 - Finalize Fit
13:38 - Create Vetiver Model & API
18:52 - Create Dockerfile
20:51 - API Deployment to Huggingface
24:03 - Using the Model
27:36 - Model Monitoring
28:58 - Improvements for Next Time
30:06 - Conclusion

#r #mlops #machinelearning #tidymodels #tidyverse #model #deployment #modelapi #api #plumber #pins #rstats #shiny #vetiver #docker #github #huggingface #datascience #ai #dataanalytics