Zaid Ajaj
Software engineer working for Pulumi, focusing on building better developer tooling for Infrastructure-as-Code and cloud native applications.
Sessions
It's 2am. Deployments are failing. Your team scrambles between ArgoCD, Kubernetes dashboards, and Git history trying to correlate what changed. What if an AI assistant could traverse these tools for you, not just suggesting commands, but actually executing investigations with your approval?
This hands-on workshop teaches you to build AI-assisted operational workflows using Agents supported by MCPs. We're going to create a multi-agent setup that leverages several MCPs (Kubernetes, ArgoCD, GitHub, Pulumi) to autonomously work on classical DevOps and SRE tasks while having a human in the loop for full enterprise compatibility.
What You'll Build
Participants will construct a working multi-agent system where:
- Specialized agents handle different operational domains (infrastructure, application delivery, source control)
- Agents autonomously detect configuration drift across Kubernetes namespaces
- Cross-agent coordination correlates ArgoCD sync failures with infrastructure changes in Pulumi
- GitHub integration queries recent commits and proposes remediation PRs
- Human approval gates maintain control over critical operations