OSS MLOps Platform

Welcome to the OSS MLOps Platform, a comprehensive suite designed to streamline your machine learning operations from experimentation to deployment.

Overview of Project Structure

• Setup Scripts

  • setup.sh: The primary script to install and configure the platform on your local machine.
  • setup.md: Detailed documentation for platform setup and testing procedures.

• Deployment Resources

  • deployment/: Contains Kubernetes deployment manifests and configurations for Infrastructure as Code (IaC) practices.

• Tutorials and Guides

  • tutorials/: A collection of resources to help you understand and utilize the platform effectively.
  • local_deployment/: A comprehensive guide for local deployment, including configuration and testing instructions.
  • gcp_quickstart/: A guide for a quickstart deployment of the platform to GCP.
  • gcp_deployment/: A guide for a production-ready deployment of the platform to GCP.
  • demo_notebooks/: A set of Jupyter notebooks showcasing example ML pipelines.
  • ray/: A guide for setting up and using Ray.

• Testing Suite

  • tests/: A suite of tests designed to ensure the platform's integrity post-deployment.

Special Instructions for Mac Users

Important Notice for Mac Users: Ensure Docker Desktop is installed on your machine, not Rancher Desktop, to avoid conflicts during the kubectl installation process. If Rancher Desktop was previously installed, please uninstall it and switch to Docker Desktop. Update your Docker context with the following command:

docker context use default

Additionally, confirm that Xcode is installed correctly to prevent potential issues:

xcode-select --install

Getting Started with a local setup

To set up the platform locally, execute the setup.sh script. For a concise setup overview, refer to the setup guide, or for a more detailed approach, consult the manual setup instructions.

Exploring Demo Examples

Dive into our demo examples to see the platform in action:

• Jupyter Notebooks (e2e):

  • Demo Wine quality ML pipeline.

  • Demo Fairness and energy monitoring pipeline.

  • Demo Ray-Kubeflow pipeline.

• Project Use-Cases (e2e):

  • Fashion-MNIST MLOPS pipeline

High-Level Architecture Overview

The following diagram illustrates the architectural design of the MLOps platform:

Key Components

• Kind: Simplifies local Kubernetes cluster setup.
• Kubernetes: The backbone container orchestrator.
• MLFlow: Manages experiment tracking and model registry.
  • PostgreSQL DB: Stores metadata for parameters and metrics.
  • MinIO: An artifact store for ML models.
• Kubeflow: Orchestrates ML workflows.
• KServe: Facilitates model deployment and serving.
• Prometheus & Grafana: Provides monitoring solutions with advanced visualization capabilities.

Support & Feedback

Join our Slack oss-mlops-platform workspace for issues, support requests or just discussing feedback.

Alternatively, feel free to use GitHub Issues for bugs, tasks or ideas to be discussed.

Contact people:

Harry Souris - harry.souris@silo.ai

Joaquin Rives - joaquin.rives@silo.ai