Crash Course: AI-Agents on ADK 2.x in Kubernetes
Mentor: Dmytro Rashko, Platform Architect, with over 25 years of experience in infrastructure and platforms
AI agents are already working, but how do you ensure they scale, stay secure, and remain under control in production?
In this 4-day intensive, you’ll explore how to build, run, and secure AI agents on Kubernetes using open-source components: Agent Substrate, kagent, agentgateway. From the first agent in a sandbox, through your own harness on ADK Go and workflows, to managed traffic, security, traces, and evals.
Mentor: Dmytro Rashko, Platform Architect, with 25+ years of experience in infrastructure and platforms.
What you’ll get as a result:
- Your own platform for AI agents on Kubernetes — skills in deployment, isolation, and efficient resource management
- Hands-on experience developing AI agents with ADK for Go and integrating them with kagent and Agent Substrate
- The ability to describe agent logic as a graph (workflows): routing, parallel branches, a human approval step
- Skills in configuring security and traffic management — routing requests to LLMs and MCP, SSO, and access control
- The ability to add your own platform extensions without forking: an MCP server, a sandbox environment, a harness
- Tools for monitoring and testing agents — end-to-end tracing, dashboards, and automated evals in CI
- Knowledge for operating the platform in production — failure diagnosis, safe component upgrades, and risk assessment
- Bonus: a set of manifests and a checklist for a pilot launch in your own team
What knowledge do you need before the course?
Required:
- Hands-on experience with Kubernetes (pods, deployments, services, networking)
- An understanding of containerization and Docker
- Basic knowledge of CI/CD and infrastructure processes
- Experience with Linux and the command line
- A laptop that runs kind (Docker)
- Your own LLM provider API key (usage costs are on your account)
Nice to have:
- Go: being able to read code is enough
- An understanding of how LLMs and AI agents work
- Familiarity with OpenTelemetry, OAuth 2.0 / OIDC, JWT
- A basic understanding of MCP and A2A
Important: The course is not aimed at beginners in DevOps or Kubernetes. It is also not for those who want to learn to build AI agents exclusively through no-code tools.
- DevOps / SRE EngineersFor those who work with Kubernetes and CI/CD and want to learn how to run AI agents in production and ensure their stability, monitoring, and safe updates.
- Platform EngineersFor those who build internal platforms and want to create their own infrastructure for AI agents with resource management, isolation, and centralized access control.
- Cloud / Infrastructure ArchitectsFor those who design infrastructure solutions and want to understand the architecture of agent platforms, their scaling, security, and integration with enterprise systems.
- Backend / Go DevelopersFor those who want to build their own AI agents with ADK for Go, integrate them with other agents and tools, and run them on Kubernetes.
Four sessions: from the first agent in a sandbox to operating the platform
day-1 Platform and your first agent in a sandboxThursday, 6:30 PM Dec 10 · 1/4 ›
- – Stack: Kubernetes, Agent Substrate, kagent, agentgateway
- – Installing the platform on kind
- – First OpenTelemetry traces
- – An agent in a sandbox
- – Automatic suspend and resume
day-2 Substrate internals, your own harness, and workflowsSaturday, 11:00 AM Dec 12 · 2/4 ›
- – How Substrate runs and places agents: snapshots, isolation classes, oversubscription
- – Harness types and your own harness on ADK Go
- – Workflows: agent logic as a graph with routing, parallel branches, and a human approval step
day-3 Managed traffic, security, and extensionsThursday, 6:30 PM Dec 17 · 3/4 ›
- – Routing models and MCP through agentgateway
- – Per-tool authorization based on a JWT claim
- – SSO, roles, TLS
- – Your own extensions without forking the platform: an MCP server, a sandbox environment, a harness
- – Client access to agents through the gateway
day-4 Observability, evals, operations, and demoSaturday, 11:00 AM Dec 19 · 4/4 ›
- – End-to-end traces in Langfuse
- – agentevals as a gate in CI
- – Platform upgrades
- – Failure injection and debugging
- – A2A between agents
- – Capstone project and final demo
The course concept and all course materials are the intellectual property of FWDAYS. Copying the content and materials of the course is permitted only with the consent of FWDAYS.
Dmytro Rashko
- – 25+ years of experience in infrastructure and platforms
- – Maintainer of the open-source projects kagent and Pi-Go
- – KubeCon speaker
Event price
Access to 4 online broadcasts (~10 hours)
Presentation and materials
Access to the course recording on the learning platform
Access to the chat in the TG, where the speaker will answer questions
Participant certificate (subject to homework completion)
Free month in Fwdays Club
Course participants receive a 10% discount on participation in Fwdays conferences
Available payment by installments from Monobank and purchase by installments from PrivatbankIf after the first lesson you realize that your expectations do not match, we will refund your money in full.