Building Governable ML Models with R
summary
Centering model work on packaging as project structure with usethis/devtools (create_package(), use_package(), document()) plus automated CI (test(), check()) yields reproducible, maintainable R models; use testthat v3 snapshot tests for deterministic training/inference checks and add simple S3 methods for predict/print so inference stubs stay concise. A two-stage model and inference pipeline stays in production for a few years and survives an Airflow-to-Databricks migration, proving the approach works in real deployments.
AI-written summary · about
chapters (7)
- 0:00Model governance introduction
- 3:55Packaging for governance
- 7:37Documentation as contract
- 9:46Testing production safety
- 14:05Legible code for maintainability
- 17:22Principles for maintainability
- 18:46Q&A
related talks
28mBuilding Developer-Centric ML Inference Platforms
57mGood Practices for Applied Machine Learning
20mBetter Care and Feeding of Machine Learning Models
39mApplying deployment oriented mindset for building Machine Learning models
41mModelPack: Packaging ML models as OCI artifacts made easy
29mHow to automate the deployment of ...- 57mGood Practices for Applied Machine Learning
05mModelpack: Standardizing The Packaging And Distribution Of...