Production Agent Operations on Kubernetes is a three-day hands-on course for operations, SRE, IT, and infrastructure staff. It prepares them to run agentic AI systems in production. The course focuses on the operational concerns around agents rather than agent application development. Participants learn how agent runtimes, model gateways, MCP servers, tool services, vector stores, queues, caches, secrets, and observability systems fit together on Kubernetes. The course emphasizes day-2 operations: deployment, monitoring, scaling, security, cost control, failure handling, upgrades, incident response, and production readiness. Through guided labs, attendees operate realistic agentic workloads and inspect traces and logs. They also enforce policies, troubleshoot failures, and build runbooks for safe production operations.
Day one labs trace a request as it moves from user to agent to tool and model gateway. Attendees then deploy a basic runtime stack using deployments, sidecars, and worker pools. They also wire in provider connectivity, API keys, and rate-limit handling. Day two shifts into observability, instrumenting workloads with OpenTelemetry traces and GenAI telemetry fields. Participants then scale workers and gateways while recovering from a simulated service failure. Security labs apply Kubernetes RBAC, workload identity, and network policy to an agent namespace. They also address risks such as prompt injection and tool misuse. Day three closes with a canary rollout and rollback exercise, a simulated incident response drill, and cost and capacity tuning. It ends with a production readiness review that ties the whole platform together.
Who Should Attend
Operations teams, SREs, IT staff, platform operators, and infrastructure teams
What Attendees Will Learn
Upon completing Production Agent Operations on Kubernetes, participants will be able to:
- Explain how AI agents, MCP servers, tools, model gateways, and supporting services run on Kubernetes
- Operate agentic workloads using Kubernetes primitives, platform services, and operational runbooks
- Monitor agent traffic, tool calls, latency, cost, resource usage, and failure modes
- Apply security controls for identities, secrets, network policy, tool access, and data protection
- Manage reliability, scaling, upgrades, rollback, and incident response for production agent systems
- Support developers and AI teams with stable operational platforms for agentic applications
- Evaluate operational readiness for agent deployments across Kubernetes environments
Prerequisites
Participants should understand basic Kubernetes concepts such as pods, deployments, services, namespaces, config maps, secrets, ingress, and logs. Prior experience with Linux operations, observability, networking, or platform operations is helpful. AI development experience is not required.