Production machine learning is everywhere. Just looking at your smartphone, nearly every popular app uses ML—recommendation engines, ETA prediction, text generation, image recognition, speech-to-text, etc.
While advances in machine learning research rightfully get a lot of credit for this growth, advances in infrastructure engineering have been just as crucial.
For early adopters, there was no playbook for production ML infrastructure. Teams were building solutions to ML-specific challenges from scratch, as in the case of Uber’s Michelangelo. But as production machine learning has become commonplace, popular architectures have started to coalesce.
Having worked with many of the engineering teams using Cortex, our open source ML infrastructure, we’ve noticed one architecture in particular becoming popular. We’ve formalized it as the MACstack

#programming #technology #software-development #machine-learning #data-science