Machine Learning Pipelines with K3s and Argo – Michael Lohan
summary
Shows how to run reproducible machine-learning pipelines on tiny clusters using k3s on Raspberry Pi (single ~100 MB binary, recommends 4 GB nodes, installs in under a minute) combined with Argo Workflows and Argo CD. Demonstrates reading student scores from a Google Cloud bucket, training a TensorFlow model in containers, deploying a web endpoint, and recovering from an injected error via Argo rollback; a full pipeline run completes in about 21 seconds to illustrate edge compute cost savings.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:46What is a container and K3s
- 3:35K3s lightweight orchestration
- 5:23Edge computing rise
- 7:14K3s production advantages
- 9:21K3s on ARM devices
- 10:27ARM architecture challenges
- 11:55Edge computing near data
- 13:39DevOps culture impact
- 15:01What is MLOps
- 16:45MLOps roles and tech
- 18:11MLOps vs DevOps goals
- 19:57ML pipeline flow
- 21:19Argo Workflows and CD
- 23:30Demo: Pipeline with Argo
- 28:14Q&A on MLOps
- 31:03DevOps and MLOps interaction
- 32:52Argo Workflows use cases
- 33:35Resources and learning
related talks
25mBuilding MLOps POCs and Sandbox Environments Using k3s and Argo
54mScaling Machine Learning Pipelines in Cloud
36mMachine learning pipeline Kubeflow, let's start
22mBuilding AutoML Pipelines With Argo Workflows and Katib
25mLeveraging Argo Workflows and Argo Events for Cloud Agnostic Enter...
07mArgo in an AI-Native World: Effective In-Cluster Machine Learning and A...
41mAlice in the world of machine learning
21mMastering the DevOps Machine Learning Pipeline