Developing and Running Machine Learning Software: Machine Learning Operations (MLOps)

Abstract

AbstractMachine learning software is fundamentally different from most other software in one important respect: it is tightly linked with data. The behavior of machine learning software is dictated as much by the data we train our models on as it is by our design choices because the information in the data is compiled into the software through the models. In a sense, models program the software automatically: developers do not completely encode its behavior in the code. Combining this idea with modern software development schools such as Agile and DevOps into MLOps has shaped how we develop and run software that incorporates probabilistic models in real‐world applications. In this article, we provide a brief overview of commonly accepted best practices for developing such software, focusing on the unique challenges that require a combination of statistical and software engineering expertise to tackle.