MLOps: best practices of putting machine learning models to production

Everyday we are being bombarded by new machine learning achievements revealed by world's leading labs in industry and academia. However, for most companies, the sad reality remains that most machine learning models never actually reach production. And for those that do, the problems in maintaining them through their lifecycle often outweigh their business benefits. This calls for a new approach how to address the complex dynamic between code, data and models that machine learning introduces. I my talk I will present the key principles of MLOps, an approach to productivization that aims to bridge the divides between data scientists, data engineers and DevOps specialists.