One of the main challenges for Kubernetes adoption in businesses is the lack of in-house skills to make the most out of a Kubernetes-based stack which is especially true for organizations implementing Machine Learning workloads developed by ML Engineers or Data Scientists whose skill set usually does not include infrastructure tooling from the cloud native ecosystem.

In this talk, you'll learn about key Kubernetes constructs, why and how to use them effectively to meet core requirements for ML workloads as we introduce Kubeflow which is an open-source project allowing users to leverage the power of Kubernetes to run the training and serving of their ML models. We will focus on Charmed Kubeflow which is a Canonical distribution of Kubeflow to show how you can leverage the power of Kubeflow to deploy and serve large machine-learning models with ease. You will also learn how to use Kubeflow's finetuning pipeline for LLM finetuning in RAG based applications with Couchbase.