Intelligence—the ability to comprehend and react—is demonstrated both as natural intelligence in humans and animals and as artificial intelligence (AI) in computers and systems. AI can deliver significant benefits to organizations by better enabling the categorization and analysis of massive amounts of data collected through mechanisms like sensors, networks, and always-connected devices.
Organizations today use AI to support progress towards their business intelligence goals and to solve problems. Machine learning (ML) is the most critical enabler of AI, employing algorithms and learning models to parse large datasets. Image and speech recognition are two examples of applications where employing ML can improve AI far faster than human analysis can. But data scientists working with ML spend an inordinate amount of time designing, configuring, and testing ML platforms, causing these highly trained specialists to spend too little time working with the data and models, which is their primary specialty. By using a process called MLOps (Machine Learning combined with Operations)—inspired by the popular DevOps model for application development—organizations investing in ML can more easily automate the continuous training and deployment of ML models at scale. MLOps adapts DevOps principles and practices to the ML workflow.
Combining development and operations in DevOps creates a more standardized and accelerated methodology for application development and deployment. Similarly, by automating the complexity and variability of the ML process, MLOPs helps lead to far more reproducible, testable, and evolvable ML environments.
Standardized, predictable, and manageable ML deployments can drive the launch of new capabilities, discoveries, and services, enabling an organization to derive more insights and value from the data it collects. These insights and values are the goal of MLOps.