Robust R Deployments: Building a Pipeline from RStudio to Production
summary
Deployment pipeline that serves live ML predictions to the Tape Cracker web app and operates successfully in production. R models sit behind a Plumber API, are Dockerized, and get built and tested via GitHub Actions with unit, integration and end‑to‑end tests. Containers progress through staging → release → production with automated integration checks at each stage. Production runs on Kubernetes with a three replicas ReplicaSet for redundancy and automatic recovery under real client load.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:43What I work on
- 2:13What is an R microservice
- 3:02DVO1 Microservice Wild
- 3:41Data Scientist vs Software Engineer
- 4:01Three DevOps Principles
- 4:36Continuous Integration Workflow
- 5:36CI Build
- 6:35Automated Tests
- 7:04Unit Tests
- 8:29Integration Tests
- 8:52EndtoEnd Tests
- 9:17Writing Tests
- 9:59Continuous Deployment
- 11:20Cluster Computing
- 12:17Replica Sets
- 13:02Conclusion
related talks
35mSimplification of Data Pipelines by Containerization
42mTrick or Treat Your Pipeline: A Haunting Tale...
39mBuilding Serverless Application Pipelines
35mProductionize your open-source project
48mHow to build solid CI-CD pipeline
37mAccelerating the Journey of an AI Algorithm to Production with OpenFaaS
26mThe Devops-Less Nightmare
06mIgnite - Blue Green Deployments Using Terraform + Packer + Jenkins