What is MLOps?¶

Machine Learning Operations (MLOps) is a practice that builds on DevOps to create an automated and streamlined workflow specifically tailored for developing, deploying, and managing machine learning models and all associated assets. Machine learning models require extra consideration for monitoring models in production and retraining if the performance declines. Planning for MLOps and implementing a strategy is a step in the AI journey if a company wants to derive greater benefit from machine learning models and manage risk. Follow us as we provide a roadmap to becoming an AI-driven company.

The engineering practice of MLOps leverages three contributing disciplines: machine learning, software engineering (especially DevOps), and data engineering. The goal of MLOps is to bridge the gap between development (Dev) and operations (Ops) and create a repeatable process for training, deploying, monitoring, and updating machine learning models.

MLOps improves the collaboration between the different stakeholders in a data science project. An AI project might include data scientists, software engineers, and subject matter experts, among others. MLOps provides tools and processes so that collaborators can contribute their part to achieve a seamless flow from model request to deploying models to solve business problems.

IBM is here to help with your MLOps journey and support you in gaining a better understanding of how MLOps works and how IBM products can be used to build a solid MLOps ecosystem for your organization. IBM's Data and AI portfolio includes the tools and capabilities you need to facilitate successful MLOps. Explore the information on this website to learn more about how to plan, execute, and customize an MLOps process that works for your needs.

MLOps with IBM Cloud Pak for Data¶

What steps and services to consider when realizing MLOps with IBMs Cloud Pak for Data?

• Follow the realizing MLOps steps in this Guide

• Follow the official documentation.

About this Repository¶

This GitHub repository has been created by several MLOps experts with expertise in IBM Cloud Pak for Data. Use this guide to learn how to design an MLOps strategy based on Cloud Pak for Data services.

What would be a example definition of MLOps?¶

MLOps (Machine Learning Operations) is a paradigm aimed at bridging the gap between development (Dev) and operations (Ops) to manage and automate the AI lifecycle. A full defintion of MLOps can be found e.g. here MLOps Paper.

Download the latest (PDF) Download the latest (DOCX)