Hi folks! We’ve been running platform engineering meetups for close to a year now (and cluminating the first year into the annual conference on 14 and 14 November - https://hasgeek.com/rootconf/2026-cfp/).

Along the way, we kept hearing the same kinds of questions, and honestly, a one-hour meetup wasn’t enough room to dig into them properly. So we decided to do workshops. Giri’s session is the first one, and we’re pretty excited about it.

https://hasgeek.com/rootconf/platform-engineering-for-production-agents/

The topic is Platform Engineering for Production AI Agents. It came straight out of what we were hearing at the meetups. Teams were trying to platformize AI agent workflows the same way they’d platformized microservices. But AI agents run inside agentic harnesses, and harnesses are a mechanism that previous platforms simply didn’t have. A harness wraps the model with state, tool execution, feedback loops, and enforceable constraints. Platformizing that is a genuinely different problem. None of the existing playbooks cover it. That gap felt worth spending a full day on.

Here’s why the gap exists. A service follows a deterministic path, so when something breaks, you usually know where to look. An agent decides its next action at runtime. That means the state of the system at any point depends on what the model decided, not what you wrote.

Giri makes this concrete by walking through five failure categories: liveness, observability, cost, safety, and correctness. These aren’t new concepts, but each one behaves differently when an AI agent is involved. To show exactly how, he runs a live incident management agent on a real system and introduces each failure on stage, one by one.

poster

Seeing a failure get introduced, diagnosed, and contained live is a lot more useful than hearing about it in theory. You get to see the actual signals, and those are the signals you’ll be looking for when it happens in your own system.

If you’re a platform engineer, SRE, or backend engineer running production infrastructure for agentic workflows, this workshop is definitely for you. No ML background needed. You just need to care about what happens when a model, not a microservice, is the system making decisions.

If that sounds like your kind of afternoon, come join us!

Date and time: Saturday, 19 September | 3 PM - 5 PM
Register here: https://hasgeek.com/rootconf/platform-engineering-for-production-agents/

Platform engineering for production agents - a hands-on workshop

{{ gettext('Draft') }}