Over the past decade, DevOps has dramatically improved software delivery outcomes for traditional software projects. Machine Learning is now mainstream and accessible enough that a similar level of maturity is becoming necessary for those projects. Thankfully, the lessons, practices, and principles of DevOps are a great basis for the emerging field of MLOps.
In this session we’ll look at how DevOps practices and principles can be applied to machine learning projects. We’ll examine the similarities and parallels between the two specialties, but also the key differences and how to adjust typical DevOps techniques to account for them. We’ll look at what’s required to construct an end-to-end MLOps solution from idea and data exploration all the way through delivery to a production system and analysis of performance.
Throughout, we’ll focus on the core principles of DevOps - delivering verifiably valuable change to end users in an efficient and robust way.