Introduction to AI/ML Toolkits with Kubeflow

Machine learning and AI are rapidly transforming the world, impacting organizations of all sizes. As executives push for AI/ML strategies, DevOps teams have been upskilling and bridging the gap between operations and development for the last several years for traditional applications. The complex machine learning application arrives just as cross-team collaboration becomes familiar. These data-dependent applications present fresh challenges for deployment and development, demanding expertise from developers and data scientists, data engineers, and machine learning engineers. How can existing engineers, with their container, Kubernetes, and cloud knowledge, navigate this terrain? Can non-engineers seeking smoother data-intensive projects find common ground with statistically-savvy data scientists? We think so! Enter Kubeflow, an open source, Kubernetes-powered toolkit that enables teams of any scale or maturity to harness the potential of machine learning. Rather than reinventing the wheel, Kubeflow simplifies the deployment of proven open-source ML systems across any cloud and even on-premise This course begins with Kubeflow, covering its origins, deployment options, individual components, and standard integrations. By the end, you'll grasp how MLOPs can ensure the successful production of ML systems, how Kubeflow opens up ML for everyone, regardless of scale, understand how to choose the ideal Kubeflow distribution for your needs so you can see Kubeflow’s "simple, portable, scalable" promise in action, and launch your own Kubeflow project. We will even touch upon some additional open source integrations so you can make Kubeflow work for you! This course caters to everyone wanting to leverage the power of machine learning. Whether you're an engineer, data scientist, or simply curious about Kubeflow, join us and discover how you can contribute to the future of machine learning! 36

• Discuss the value of MLOPs for production systems and how it relates to DevOps
• Recognize common machine learning platform patterns and the problems they seek to solve
• Explain the model development lifecycle
• Define and identify common machine learning frameworks
• Discuss the value proposition and goal of the universal training operator
• Research and select a Kubeflow distribution based on your needs or, at the very least, have an informed conversation with a vendor.
• Launch and leverage a Kubeflow Notebook.
• Launch a primary Kubeflow pipeline.
• Discuss additional popular Kubeflow integrations.
• Familiarize yourself with Katib and Hyperparameter tuning

• Experience with cloud computing
• Familiarity with DevOps and cloud native principles
• Basic programming experience
• Experience with technical documentation
• Experience with open source projects in general.
• Basic understanding of Kubernetes might be helpful but not necessary