robust-r-deployments-building-pipeline.md

talk_id	talk_slug	talk_type	talk_tags	session_slug	sched_url	talk_title	talk_title_short	talk_materials_url	speakers
22198	robust-r-deployments-building-pipeline	regular	html/css/js interop production	data-science-in-production	https://rstudioconf2022.sched.com/event/11iYu	Robust R Deployments: Building a Pipeline from RStudio to Production	Robust R Deployments		name affiliation url username photo bio David Maguire dv01 webpage twitter github linkedin affiliation https://github.com/davemaguire https://www.linkedin.com/in/david-maguire-30636956/ https://www.dv01.co/offerings/tape-cracker/ david_maguire /assets/img/2022Conf/_talks/22198_david-maguire.jpg A physical scientist by training, David leverages the scientific method along with the statistical capabilities of R to solve business issues. He has applied this skillset to a variety of industries, including pharmaceutical manufacturing and, currently, structured finance at dv01. After witnessing non-robust deployments prevent excellent models from creating value, David focused on creating robust and scalable model pipelines in the cloud. At his current position at dv01, David manages a customer facing plumber API. This API is the entry point to one of dv01's top products, Tape Cracker.

R is often maligned as a poor fit for production deployment systems. At dv01 we deploy Plumber API that serves machine learning models to Tape Cracker, a client facing web application. With R in production we use the same codebase throughout the data science pipeline, saving time and resources while reducing the chance for bugs in the hand-off. To accomplish this we use industry standard continuous integration and deployment tools to deploy our API to compute clusters in the cloud. Our pipeline progresses stepwise through staging, release and production environments. Automated integration testing at each step enables a robust and reliable deployment. In this talk R users will learn strategies to deploy R code in production environments.