Data Science in Production: The Way to a Centralized Infrastructure
summary
Covestro centralizes R and Python work with eoda and Posit into a replicable, staged platform: Posit Workbench on AWS with Kubernetes offhost sessions (low/medium/high resource tiers), NFS-shared home and app directories, externalized package manager and metadata, and Posit Connect split into dev/test and prod. Implementation follows assessment → MVP iterations → operation, built with Terraform IaC. Outcome: improved collaboration, maintainability, performance and compliance; even a 25‑year R user abandons the laptop.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:53Covestro case study
- 3:20Centralized infrastructure challenges
- 5:06Three-step infrastructure setup
- 7:13Covestro’s goals
- 10:25Posit-centric design
- 13:56Posit Connect workflow
- 16:28Additional components
- 17:10Infrastructure advantages
- 18:55Lessons learned
- 19:57Closing & Q&A
related talks
18mData Science in Production Has Never Been So Easy
05mCode-First Data Science
20mChampioning modern science workflows to benefit dairy farmers
58mUpdates from Posit
21mEnd-to-end data science with the Positron IDE
01h 08mGet the Latest on Posit's Commercial Products
36mBuilding a Scalable Data Science & Machine Learning cloud using Kubernetes
20mGiving your scientific computing environment (SCE) a voice