HOW TO BUILD YOUR OWN AI AGENT: THE COMPLETE ARCHITECTURE FOR ENGINEERING TEAMS | AI SRE SUMMIT 2026

The gap between handing your team a general-purpose coding assistant and giving them an agent that truly understands your company is enormous. It's the difference between a brilliant intern who knows nothing about your business and a senior engineer who's been with you for years. Generic tools don't know your internal systems, your runbooks, your deployment patterns, or your infrastructure quirks. Every engineer ends up reinventing the wheel, and knowledge stays trapped in individual heads or buried Slack threads. in this talk, we l walk through the complete architecture for building your own Al agent. Not theory - the reall decisions you'll face and how to make them. We'll cover every component: system context that teaches the LLM how your company operates, tools that let the agent take action, vector databases and knowledge graphs for semantic search over internal knowledge, memory so the agent learns from its own experience, multi-agent orchestration with specialized agents instead of one monolithic one, MCP integration so engineers use their existing tools, security with guardrails and human-in-the-loop for critical operations, observability for a system where inputs, processing, and outputs are all non-deterministic, and model routing to keep costs under control. This session was presented at the 2026 AI SRE Summit by Viktor Farcic, DevRel at Upbound