The scenario
PipelineSmith solves the copy-pasted-Jenkinsfile problem for ML teams. You build a Jenkins shared library with stages for data checks, training, evaluation gates, image builds and Kubernetes deployment, then onboard two sample ML projects to it.
System architecture
01 Git push Any ML repo
02 Shared library Standard stages
03 Train + evaluate MLflow tracking
04 Build image Docker registry
05 Kubernetes Rolling deploy
High-level flow of PipelineSmith, from input to output.
What you will learn
- Jenkins shared libraries
- Evaluation gates in CI
- Deploying models to Kubernetes
What you will build
- Shared library
- Two onboarded ML repos
- Model deployment to a cluster
Build plan
- Jenkins setup
- Library stages
- Eval gate
- K8s deploy
- Onboard repos