A former startup CEO and technical founder, Luke has recently started a consultancy specialising in MLOps. He offers clients the ability to take advantage of the learnings from his former MLOps platform company, Dotscience, to help them develop their MLOps capabilities, integrating with platforms such as Kubeflow and MLflow.
GitOps for MLOps: Going Beyond Code and Environment Versioning
GitOps is great for versioning code and environments (containers) and the interplay between them in production for software DevOps. But MLOps (DevOps for Machine Learning) is significantly more complex, demanding versioned data and models as well. In this talk, Luke will explore the differences between DevOps and MLOps and look at what needs to be done to capture the additional complexity in a production MLOps system. He hopes to start a conversation about how tools like Flux can be extended to better support MLOps workflows.