Workshop
Productionizing agentic workloads (Sunnyvale)
Tuesday December 1, 2026 · 9:30 AM to 4:30 PM Pacific
Google, 1212 Bordeaux Dr, Sunnyvale, CA
Most engineering orgs have “adopted” AI. The question is how do you start to productionize both internal and external workflows that include inference to make them more reliable - whether that’s for customers, a software factory or a more general agent factory?
The answer requires both the right strategies and the right tactics, so I’ll be hosting a full day workshop at Google in Sunnyvale on Tuesday December 1st bringing together engineering leaders and SREs to dig into both.
If you lead engineering
This is a chance to connect with and learn from engineering leaders at other companies as we plan for 2027. We’ll be sharing what is and isn’t working for agentic engineering and will look at both the architecture for composing a reliable software/agent factory and the steps to drive adoption incrementally and safely.
I’ve interviewed hundreds of engineering leaders for my upcoming O’Reilly book on Scaling AI Adoption in Engineering so I’ll share what I’ve learned and we’ll be breaking out into small groups to discuss our challenges and practical next steps - from driving adoption and justifying token budgets to understanding the infrastructure required to make agentic workloads both safe and reliable.
Bring along a IC from your platform/SRE org for the technical breakout!
If you’re an IC (platform/SRE)
Ever been given an agent and asked to productionize it? We’ll be taking on an “agent reliability challenge”. It’s a really well structured lab where we start with a heterogenous mix of agents (LangGraph, CrewAI and Google ADK agents), instrument them and use the traces to identify failures, fix the issues and then turn the failures into evals for future runs.
It’s a chance to see good practices for observing, fixing, evaluating and tuning agents in production and a great chance to compare notes with other SREs and platform engineers so you can share what is and isn’t working at your companies for managing agentic workflows.
The day
We’ll all spend part of the morning working in a team on three different agents (LangGraph, CrewAI and Google ADK, on Google Cloud) to see good practices for instrumenting them and identifying issues using traces.
We’ll then break out folks who want to talk adoption and architecture from those who want to dig way deeper to get hands on experience of observing, fixing, evaluating and tuning agents in production.
The logistics
Tuesday December 1, 9:30 AM to 4:30 PM, at Google, 1212 Bordeaux Dr, Sunnyvale, CA 94089. Free. Lunch and snacks are provided. Please bring a laptop for connecting to the lab.
Hope to see you there - should be some great discussions and learning!
Peter Bell
Founder/CTO
Gather.dev