Modern machine learning architectures: Data and hardware and platform, oh my

The move toward adopting machine learning in modern systems is not quite the same as simply deploying a new feature or framework. It requires disciplined thinking about where to place your data, where to place your analysis, where to place your models, and more. The choice of language and implementation increasingly have implications on the properties of your runtime system. Many of the most interesting trends in architecture these days fundamentally come down to finding cost-effective ways of doing what you need to do computationally. Machine learning systems are no different and will benefit from these developments as well.

Brian Sletten takes a deep dive into the intersection of data, models, hardware, language, and architecture as it relates to machine learning systems in particular, but the overall industry in general.