Ten Things to Know About ModelOps

Chapter 2. Ten Things to Know About ModelOps

1. ModelOps Is the Enterprise “Operating System” for AI/ML Algorithms

While AI/ML and deep learning methods get the most attention in the press, blogs, and on the web, the reality is that for the majority of applications the model authoring may be the simplest part. The supporting logic, processing, deployment, and controls—and in particular setting up an efficient delivery system of solutions infused by AI/ML—are the much bigger challenge.

ModelOps, therefore, is a kind of operating system for AI/ML models. That is, it creates a platform and set of tools to deploy algorithms in a reliable, repeatable, compliant, safe, and efficient way. Like the operating system on a laptop or mobile phone, it provides access to the right algorithm and to the right place at the right time when it is needed.

Many Types of Models

To reiterate, models are just transformations of data, in order to extract useful information. There are many types of models that can convey useful information, help make better decisions, or drive effective automation. Continuing with the operating system metaphor, it is a mistake to limit ModelOps to just one “app”: ModelOps tools must deploy AI algorithms built on any platform or provided by any cloud service.

Any statistical analysis, aggregation, set of rules, summary, or BI dashboard represents information, abstractions, and inferences from observed data that guide strategies and decisions. All of the above are “models,” ...