AI and Analytics at Scale

Preface

Getting large-scale, data-driven applications for AI and analytics into production does not have to be as challenging as many people make it. There are fundamental aspects of design and data infrastructure that are often overlooked, aspects that can make the difference between failed systems and reliable, successful production systems, even at very large scale. You should, for instance, be able to scale data, the number of applications, and the teams who build them without having to scale your IT team.

In this report, we bring you an up-to-date look at what works and what does not, based especially on our observations from the past several years working with a wide range of businesses with successful production systems at scale. This report is not intended to be an encyclopedic guide to every aspect of AI and analytics at scale. Instead, we want to highlight some of the practices that have been particularly beneficial, especially those that are often overlooked.

You’ll see common themes among our recommendations for how to handle AI and analytics at scale while also keeping flexibility to increase scale and pivot to new applications or technologies as desired. These high-level concepts include developing a comprehensive data strategy that spans your organization and taking advantage of the isolation, flexibility, and predictability that containerization of applications affords. We describe capabilities of and approaches to building data infrastructure that have allowed ...