Building AI/ML Pipelines on Kubernetes

Learn how to build AI/ML pipelines on Kubernetes in this conference talk featuring platform engineers from Google, LanceDB, and Uber who discuss the evolution from waterfall development to data-driven, multi-path AI workflows. Discover how Kubernetes has become the standard container orchestration platform for AI/ML use cases including extracting insights, parsing unstructured data, and handling missing multimodal data across images, text, audio, and video. Explore the challenges platform engineers face when building data pipelines that support multi-model infrastructure, enabling AI applications to dynamically select the optimal model for each request. Gain insights into constructing data pipelines in Kubernetes specifically designed for multimodal AI that optimizes data retrieval, advanced reasoning, and generation capabilities from experienced practitioners working at scale in production environments.