In the talk I show how to make thoughtful ML deployments on Kubernetes. With thoughtful ML deployments I focus on the challenge to keep ML: manageable, accountable and explainable. I propose 5 design principles to achieve this.
The goal of the talk is to show developers that want to move from just creating models to productionizing them using proven best practices to get started. Pointing them both to promising technologies and challenges in a real world ML deployments. While pointing to parallels and differences from proven DevOps practices in traditional software engineering.