I lead the AI Infrastructure and Applied Engineering Developer Relations teams at Google Cloud. The first team covers GKE and the accelerators that power large-scale AI training, fine-tuning, and inference. The second puts Google Cloud engineers alongside a customer's own team to build a mission-critical system together. Together they help enterprises get the most out of the world's most capable AI hardware and run agents in production with confidence.
Every resilience pattern you know assumes the machine shows up. Accelerator scarcity breaks that assumption, and the pattern catalogs have not caught up. Here is the Compute Fallback Ladder—an ordered set of rungs your workload can run on, a selector that picks the highest obtainable one, and a promotion path back up when capacity returns.
Agent density is a lifecycle problem, not an isolation problem. Google's own GKE Agent Sandbox benchmark puts the sandbox at 44% and suspend and resume at up to 3.5x—the sandbox is the smaller half. And upstream Kubernetes 1.37 is quietly turning checkpoint and restore into a kubelet primitive. Here's why the economics live in the lifecycle, not the box.
I'm hiring a Senior Developer Relations Engineer for GKE and AI Infrastructure. Instead of describing the work, I built it: a small, playable cluster scheduler that teaches the real problem behind running AI at scale—gang scheduling, memory ceilings, spot preemption, and a TPU slice with actual topology.
The finale of the series: the mature end-state the first three parts build toward. When every server-side capability's trust boundary is a decision on record—the deployment tier you chose, the toggle you set, the contract you signed—saying yes to another team stops being a risk assessment and becomes a lookup. Fine-grained capability governance is what buys you safe expansion. Maturity, not lock-in. Boring, in the way a well-run system is boring.
Parts 1 and 2 argued that a capability's trust boundary is a separate decision from the model's. Part 3 is the runbook that makes the decision stick: the org-policy constraints that gate partner web search and structured outputs, set once at the organization tier by gcloud and Terraform, plus the seams org-policy doesn't reach—VPC Service Controls, request-response logging, the grounding-provider choice. Which toggle, at which scope, with deny-wins precedence, so a good-faith developer can't trip a data path nobody chose.
On February 23, 2026, Google quietly made the retention terms for Grounding with Google Search better: the window dropped from thirty days to up to three, and what's kept narrowed from your prompts and output to short-lived debug logs. The old thirty-day, prompts-and-output language didn't vanish—it now describes Grounding with Google Maps, a sibling capability under the same terms. Here's what the corrected terms actually say, how to verify the date yourself from Google's dated archive, and the durable lesson: govern facts that move—cite the dated primary source, hedge the number, and don't hard-code a digit that will drift out from under you.
The builder's companion to Governed Growth—the same reference repo from the implementer's chair. Why I built a one-tool MCP server to hand an assistant a reliable, governed Google search, how Gemini grounding and citation extraction work in the code, and how a full, human-validated eval picked the default model. The surprise: gemini-3.1-flash-lite matched the larger models on quality, grounded most reliably, and cost the least.
Opening a platform to more teams and more models—first-party and partner, such as Gemini and Claude—quietly assumes those capabilities share one trust boundary. They don't. Multi-model expansion is a governance-maturity story: Google Cloud lets you grant capability broadly while keeping each model's data path a deliberate, fine-grained decision. You can grow usage without growing exposure.
GitHub's REST and GraphQL APIs have separate hourly rate limits—two budgets, not one. Most automation hammers a single bucket while the other sits full. Here's how to see both, why agents drain GraphQL fast, and how routing work across both effectively doubles your headroom.
Traditional SDLC fails for autonomous AI. The Agentic Manifesto introduces the Agentic Delivery Lifecycle (ADLC)—a new operating model for engineering, governing, and tuning non-deterministic systems at scale.
A deep-dive playbook for building scalable, enterprise-grade AI agents. Learn how to move beyond monolithic designs by decoupling tools with a powerful microservices architecture using ADK, the Model Context Protocol (MCP), and Google Cloud Run.
Explore the evolution from basic LLM interactions to crafting sophisticated System Instructions, and discover how meta-prompting—using AI to refine AI prompts—can unlock more powerful and collaborative AI agents. This post details the journey and a practical approach to building AI that helps you build better AI.
Stop wrestling with Python versions and dependencies. Learn how to build a fast, simple, and unified development environment using uv
