Kubeflow for Poets · Part 1 of 8
Kubeflow for Poets: A Guide to Containerization of the Machine Learning Production Pipeline
27 March 2019Updated 14 October 20261 min read
This series shows how to run a machine learning production pipeline as containerized components with Docker, Kubernetes and Kubeflow on Google Cloud. This overview introduces the series and links to its eight parts.
This series gives a systematic approach to putting machine learning pipelines into production with Kubeflow on Kubernetes. Building a model is one piece of a larger system of tasks and processes that together deliver a machine learning product. Kubeflow applies the microservices pattern to machine learning. Each stage of the pipeline runs as a separate container. Kubernetes orchestrates the containers.
Kubernetes runs on any platform. This series deploys a machine learning product on Google Cloud. It uses BigQuery to stage data, Dataflow to transform data and Gemini Enterprise Agent Platform to train and serve models. Kubeflow runs on a Google Kubernetes Engine (GKE) cluster and orchestrates these steps.
The series first used Cloud Machine Learning Engine to train and serve models. Google later renamed that service AI Platform, deprecated it and moved its features to Vertex AI. In April 2026, Vertex AI became part of Gemini Enterprise Agent Platform.
Contents
The series has eight parts:
- Overview (this page)
- Introduction
- Microservices architecture
- Docker
- Kubernetes
- Kubeflow
- Kubeflow Pipelines
- Deploying an end-to-end machine learning solution on Kubeflow Pipelines
Links
- Docker
- Kubernetes
- Kubeflow
- Kubeflow Pipelines
- BigQuery
- Dataflow
- Gemini Enterprise Agent Platform
- Google Kubernetes Engine
Source Code and Contributions
The source code for the series is on GitHub. Contributions and corrections are welcome as pull requests.
Ekaba Bisong is a Ph.D. researcher at the University of Victoria and the founder of SiliconBlast. More about me