Overview
Explore the integration of machine learning models in streaming applications with this insightful conference talk by Boris Lublinsky, Principal Architect at Lightbend. Delve into the shift from batch to real-time streaming and its impact on user expectations. Discover how machine learning enhances mission-critical real-time applications, enabling innovative solutions and improving traditional scenarios like fraud detection and predictive maintenance. Learn about two main approaches to model usage in streaming applications: embedded models and external servers. Gain hands-on knowledge of Cloudflow, a specialized streaming framework for developing, orchestrating, and operating distributed streaming applications on Kubernetes. Understand how Cloudflow supports the entire application lifecycle, from development to operation. Examine the implementation of different model serving approaches using Cloudflow and its integration with Kubeflow. Conclude with valuable recommendations for choosing the most suitable model serving approach based on specific requirements.
Syllabus
Boris Lublinsky - Using Model Serving in Streaming Applications
Taught by
Toronto Machine Learning Series (TMLS)