As AI and automation reshape modern software, datasets themselves can become active, intelligent components in cloud-native systems. This session introduces Agentic Datasets - a Kubernetes-based architecture where datasets act as self-managing data products capable of reasoning, triggering workflows, and coordinating with AI models.
Each dataset runs as a DatasetAgent, defined as a Kubernetes Custom Resource (CRD) that encapsulates metadata, logic, and declarative workflows packaged as OCI bundles. Within the cluster, these workflows execute securely as Jobs or Pods, performing tasks such as data validation, reindexing, or embedding generation - all without external orchestration.
We’ll explore how controllers, service meshes, and GitOps pipelines enable these agents to interact safely and autonomously, and how LLM-powered reasoning can enhance decision-making within workflows.
Attendees will learn how to build and deploy their own intelligent, AI-assisted data products on Kubernetes using open-source components, bridging the gap between DevOps, MLOps, and intelligent automation.