CHAPTERS
Why this talk: bridging AI coding tools and Google Cloud deployment
The speaker opens by surveying the audience’s use of AI coding tools and highlights a gap: fewer people use those tools to actually build and deploy on Google Cloud. He frames the session as a practical walkthrough to close that gap using Claude on Google Cloud.
The enterprise software team—and how Claude Code augments each role
The talk outlines a typical enterprise product team (PM, UI/UX, software engineer, security engineer, analytics) and explains that Claude Code can augment work across the entire lifecycle. The speaker sets up the main narrative device: he’ll wear five “hats” and build a feedback app live.
Who’s speaking: Ivan Nardini’s role and the Google Cloud + Anthropic partnership
Ivan Nardini introduces himself as a Google Cloud Developer Advocate working in partnership with Anthropic. He emphasizes his focus on enabling developers to build and deploy apps on Google Cloud using Claude.
Getting started: configuring Claude Code to use Claude models on Google Cloud
Ivan explains how to set up Claude Code to run Claude models on Google Cloud, focusing on a straightforward flow using Application Default Credentials (ADC). A setup wizard detects project/region, verifies model availability, and allows pinning models for use in sessions.
Why run Claude on Google Cloud: cost, security, availability, and governance
The speaker lists practical advantages of using Claude models via Google Cloud, including pay-per-token pricing, scalable capacity, and reduced operational overhead. He also highlights enterprise controls: data stays within your project, policies are enforceable, and endpoints support multi-region availability.
PM hat demo: from napkin sketch to wireframe + GitHub PR
As the “PM,” Ivan shows how Claude Code can transform a rough sketch into a usable wireframe quickly. He emphasizes that even non-technical roles can contribute by having Claude generate assets and open a pull request to share with the team.
UI/UX hat demo: using Planning Mode to turn wireframes into production UI pages
Switching to the UI/UX role, Ivan uses Claude Code’s Planning Mode to propose a build plan before coding. He creates a more production-ready interface with multiple pages (landing, thank-you, and a dashboard view) and then generates code and submits another PR.
Software engineer hat: architecture design with Developer Knowledge API + skills
Ivan introduces Google Cloud’s integration with the Claude ecosystem: official Google Cloud skills and the Developer Knowledge API with an MCP server. Claude can fetch fresh documentation and guides (refreshed every 24 hours) to help design an architecture: Cloud Run API, Firestore storage, BigQuery ingestion, and Looker visualization.
Parallel build with subagents: API, data pipeline, and dashboard built like a sprint
To simulate a team sprint, Ivan uses Claude Code subagents to parallelize tasks across components (API, BigQuery pipeline, dashboard). Claude implements the system, runs tests, and prepares deployment assets for Google Cloud.
CI/CD to Google Cloud: Cloud Build + Cloud Deploy and dev environment rollout
Claude generates the infrastructure and pipelines needed for deployment, using Cloud Build for CI and Cloud Deploy for CD. After merging the PR, a workflow triggers builds and releases, deploying automatically to a development environment.
Security engineer hat: custom security review plugin, input validation, and least privilege
Before production, Ivan demonstrates a security review using a custom Claude Code plugin to reflect real organizational requirements. The review checks input validation and tightens service account permissions; Claude updates code, re-tests, and opens a PR for the fixes.
Promotion to production: gated approval using Cloud Deploy
After security changes are merged, the pipeline runs again and produces a new release. Ivan highlights how Cloud Deploy supports review and approval gates: deploy to dev, validate, then promote to production through an explicit approval step.
Analytics & insights hat: BigQuery + Looker dashboards via MCP servers
Finally, Ivan shows how to collect and analyze usage metrics (e.g., time-to-submit feedback) to improve UX. Using MCP servers for BigQuery and Looker, Claude can compute statistics from raw data and generate a shareable Looker dashboard link suitable for PM stakeholders.
Wrap-up: Claude Code capabilities + Google Cloud hosting make end-to-end delivery smoother
Ivan summarizes the combined value: Claude Code features (Planning Mode, subagents, MCP, skills, plugins) accelerate prototyping through deployment and analytics. Google Cloud provides frictionless access to Claude models plus robust hosting and operational standards, and he points viewers to repo/quickstarts/docs for next steps.
