From Notebook to Kubeflow Pipelines to KFServing - The Data Science Odyssey

Explore a comprehensive tutorial on leveraging Kubeflow for data science and machine learning workflows. Learn to deploy Jupyter Notebooks as Kubeflow pipelines using Kale, optimize model training with Katib for hyperparameter tuning, and serve models using KFServing. Discover techniques for running thousands of pipeline iterations with caching and garbage collection, while tracking and reproducing pipeline steps along with their state and artifacts. Gain hands-on experience with MiniKF deployment, pipeline creation, and notebook management. Dive into advanced topics such as manual pipeline compilation, experiment visualization, and model serving through KFServing API. Perfect for both data scientists seeking an intuitive GUI-based approach and ML engineers looking to build advanced, reproducible workflows.

Introduction
Agenda
What is Kubeflow
Use Cases
User Interface
Command Line
ML Code
Highlevel overview
The agenda
Deployment
Creating a MiniKF
Pipelines
Benefits
Workflow
Beta Management
Notebook Overview
Annotations
Compile Run
Analyze Notebook
Python Libraries
Manual Pipelines
Compile and Run
Catib UI
Experiments Graph
Cached Steps
Visualizing Steps
KFServing
Restoring a Notebook
Restoring a Notebook with the State
Restoring a Notebook with the Server
Restoring a Notebook from a Snapshot
KFServing API
Preprocessing
Models UI
Summary
Community