Building MLOps POCs and Sandbox Environments Using k3s and Argo
summary
End-to-end MLOps POC that runs on a single-node k3s cluster and uses Argo Workflows to run ETL → training → model upload and Argo CD to deploy a Helm chart exposing an inference API; pipeline reads a CSV of student scores, uses pandas/scikit-learn to train a model, stores artifacts in a Google Cloud bucket, and produces a prediction CSV. Demo shows successful runs, Helm-based deployment, ingress endpoint testing and a history rollback. k3s single-binary lightweight Kubernetes Argo Workflows + Argo CD orchestrate ETL→train→deploy
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:54Basic Concepts
- 4:39Difference between devops and envelopes
- 12:05Demo
- 13:44Argo UI
- 15:55Workflows
- 17:55Testing
- 18:32Deploy
- 20:08Rollback
- 21:09Pipelines
- 21:59Python Code
- 23:31Inference
related talks
38mMachine Learning Pipelines with K3s and Argo – Michael Lohan
38mMachine Learning Pipelines with K3s and Argo – Sergio Mendez
27mBuilding Secure MLOps Pipelines With KitOps + Argo Workflows
28mScaling MLOps on Kubernetes with Amazon SageMaker Operators
22mBuilding a MLOps Platform with Kubeflow on GKE
25mLeveraging Argo Workflows and Argo Events for Cloud Agnostic Enter...
24mDemo of Kubernetes at the Edge for AI Model Deployments
39mBuild and Deploy Your Application with Argo on Kubernetes