Introduction to KubeFlow: Using and Use Cases

Explore the fundamentals of KubeFlow and its applications in machine learning with Kubernetes in this informative conference talk. Gain insights into container technology, Kubernetes' desired state management, and the core components of KubeFlow. Learn about collaborative filtering, rating matrices, and the use of Minio for centralized storage. Discover how to implement machine learning models using Jupiter notebooks, convert them to MLJobs, and deploy them for machine serving. Follow a practical demonstration of a recommender system for product suggestions based on customer purchase history. Delve into advanced topics such as concept drift, continuous model updates, and custom components like Argo workflow. Get hands-on experience with code samples and learn how to engage with the KubeFlow community. Master the essentials of KubeFlow, machine learning, and their integration with Kubernetes for efficient deployment of complex workloads across private and public clouds.

Intro
What are containers?
What is kubernetes
Kubernetes desired state management
What is machine learning?
What is kubeflow?
Kubeflow components
What's new in 0.6?
Meet Kubeflow
Cluster view
Motivating example
Collaborative filtering
Rating Matrix
Using Minio as a centralized storage
Kubeflow options for machine learning and model serving
Using Jupiter for creating implementation
Converting implementation to TF Job
Running TF Job
Deploying TF-serving
Using TF Serving in streaming applications
Concept drift
Continuous model updates implementation
Additional custom components
Argo workflow
Bringing it all together
Try this yourself