Kubernetes* Module: Deploy Cloud-Native, AI Workloads on AWS*

Machine learning applications continue to get larger and more complex, which further underscores the importance of scalability and effectively deploying and operationalizing AI pipelines.

This session introduces a way to do exactly that. It presents Intel’s first cloud optimization module for Kubernetes*, an end-to-end, open source reference architecture designed to help developers build Intel® Xeon® processors that are based on Kubernetes clusters with automated scaling to manage high volumes of training requests.

Topics covered include:

• An overview of the new cloud-optimization architecture.
• Using Kubernetes to deploy and operationalize AI on Amazon Web Services (AWS)* clusters that are based on Intel Xeon processors.
• Managing high volumes of training requests with Kubernetes, load balancer, and auto scaler.
• Handling inference requests to the machine learning prediction service using serverless lambda functions.
• Demonstrations that include how to integrate the Load Default Risk Prediction reference solution into your AI pipeline and how to accelerate containerized training and inference of an XGBoost classifier

Skill level: Intermediate

This session include the following tools, which are free to download:

• Intel® Optimization for XGBoost* as a stand-alone version or as part of the AI Tools
• daal4py, which is part of the stand-alone version of Intel® oneAPI Data Analytics Library or as part of the Intel® oneAPI Base Toolkit
• Intel® Extension for Scikit-learn* as a stand-alone version or as part of the AI Tools

• Intel Device Plug-ins for Kubernetes: Shows how to use the optimized matrix multiplication routines for Intel® oneAPI Math Kernel Library.
• XGBoost Get Started Sample
• Intel Extension for Scikit-learn Get Started Sample