In this presentation, we will look at the current state of industrialization of Machine Learning solutions. Algorithms (even extremely sophisticated ones that can develop themselves) ultimately make up only a small part of each Machine Learning system. Successfully setting up such a system requires a huge architectural and engineering effort. We will look at what makes Machine Learning inherently difficult to do at an industrial scale and then explore recent developments that have made it much easier to do so. Some key talking points will include designing modular workflows; supporting those workflows with a microservices approach; and prototyping at scale.
Karl Schriek, Head of AI / Leading Machine Learning Engineer, Alexander Thamm GmbH