Put R in prod hex logo

Tools and guides to put R models into production

Production

/prəˈdəkSH(ə)n/
noun
The systems that continuously run code to help an organization: "the recommendation engine is in production and customers are using it"

R

/är/
noun
An open source language for statistics, visualizations, and machine learning: "I trained a deep learning model in R"

Talks

Push straight to prod: API development with R and TensorFlow at T-Mobile

An rstudio::conf 2019 talk on how T-Mobile uses R in production

This presentation covers the work done at T-Mobile to use R and Keras to develop machine learning models for natural language processing. It show how the successful deployment of R has changed the customer experience.

We’re hitting R a million times a day so we made a talk about it

An rstudio::conf 2020 talk on lessons leard from a year of R in production

After having R in production for a year at T-Mobile, the R models were being hit a million time a day. This talks covers the lessons learned from scaling R up in production in an enterprise setting.

Build your Machine Learning Datasets with AWS SageMaker Ground Truth

A tutorial on using Ground Truth to label training datasets

Successful machine learning models are built on the foundation of large volumes of high-quality training data. Amazon SageMaker Ground Truth significantly reduces the effort required to create datasets for training. Learn how companies like T-Mobile are using Amazon SageMaker Ground Truth.

Rn’t u glad u put R in prod

A walkthrough on how plumber and Docker can be used for R

This talk covers taking a model in R and putting it into production with plumber and Docker. It uses the TensorFlow model from the talk Deep learning isn't hard, I promise.


Code

R TensorFlow API Docker container

Everything you need to create a Docker container for an R neural network

This repository contains R code, Dockerfiles, and configurations needed to run your own neural network as an API with HTTPS protocol.

Loadtest

Load test your APIs from R

This R package helps you test your APIs to make sure they can handle many users at once. Run a load test with one line of code!

Example project: pet names

Make a Docker container that generates pet names

This full example generates pet names using a TensorFlow model, and includes the plumber API and Dockerfile code. Used in the R NYC Conference talks: Deep learning isn't hard, I promise and Rn't u glad u put R in prod.


Blog posts

R can API and so can you!

USING PLUMBER TO CREATE AN API IN R

This blog post covers how to use the plumber package of R to create a RESTful API. With this, you can use R code to start a microservice allowing other people to call your R code as needed.

Using Docker to deploy an R plumber API

GET YOUR R CODE TO RUN ANYWHERE AS A SERVICE

The second blog post of the series works through how to take an web service created in R with plumber and put it in a Docker container.

Enterprise web services with neural networks using R and TensorFlow

DEPLOYING A NEURAL NETWORK INTO PRODUCTION ENVIRONMENTS USING AN R BASED STACK

The final blog post discusses the T-Mobile open source Docker container that contains all the things you need to train a neural network and to deploy it in an enterprise environment.

Small data, big value

How simple and quick machine learning projects can beat out massive technology investments

Machine learning models can be used in small ways to provide simple victories for companies that can quickly produce results with lower risks and lower costs compared to large ML projects.


Demos

Are you a customer machine learning model

A Shiny demo showing an R TensorFlow model in action

This demo, as described in the Small data, big value blog post, shows a deep learning model used in production at T-Mobile. This demo is deployed using the open source R TensorFlow API Docker container.