Urgent Requirement II Gemini AI Architect | San Ramon ,CA. (Hybrid 3 day onsite)

About the Role

You are the architect for the Governance Control Tower — the single person who owns its design end to end and who the platform leadership deals with directly.

You will define the architecture standards and the checklist every agent must clear before it goes live, run the review board relationship , and lead a cross-functional delivery pod that implements and then operates the platform. You stay hands-on. This role sets the standard and then proves it works — designing the reusable agent layer, resolving the identity and retrieval questions that determine whether the governance model is real or decorative, and carrying the technical credibility.

You will lead the pod through two phases: build, standing the Control Tower up; then run, where a right-sized support pod takes over day-to-day operation and you remain the architectural owner.

The Platform Environment

• Gemini Enterprise — the employee-facing surface: managed chat, enterprise search, agent gallery, and app-scoped experiences serving all business functions.
• Agent Platform — the build, govern and operate layer: ADK and agent runtime, agent and tool registry, agent identity, agent security, agent observability.
• Custom front-end applications calling Gemini Enterprise via API, with end-user identity propagation.
• Integration and data layer — enterprise connectors across Google Workspace and Microsoft 365, BYO and vendor-managed MCP servers, and federated retrieval across function-scoped data stores.

What You'll Own

The Control Tower is delivered as five workstreams. You own the architecture across all five.

Governance & Standards

Registry & Gateway Operations 

Connector & Retrieval Engineering 

Identity & Entitlement Enforcement 

Observability & Cost Control 

Across all of it, you will also:

• Design the reusable top-level agent layer — orchestrator, context engineering, retrieval, synthesis and response — that acts as the common entry point for every application on the platform, so that query planning, grounding, re-ranking and citation behave identically wherever a user enters.
• Lead the pod: set technical direction, review the work, and be accountable for what ships.
• Translate between the platform leadership and the delivery team, turning direction into architecture and architecture into a defensible plan.
• Hold the quality bar — evaluation datasets, threshold gates, regression testing — so the platform stays reliable as agent count scales.

What We're Looking For

• 10+ years in software engineering and architecture, with 3+ years designing and running applied AI systems in production.
• Proven experience as the architectural owner of an enterprise platform — you have set standards that other teams had to follow, and made them stick.
• Hands-on with Google Gemini Enterprise and ADK, or a directly comparable enterprise agent platform, including agent runtime, registration, identity and observability.
• Deep experience with multi-agent systems in production — orchestration, routing, tool use, memory, human-in-the-loop — with real operational ownership rather than prototype work.
• Strong grounding in RAG and retrieval architecture: vector stores, embedding models, chunking strategy, hybrid search, and the difference between retrieval that works in a demo and retrieval that works across a messy enterprise estate.
• Identity and access depth. OAuth2, SAML, RBAC, token exchange, service-account versus end-user credential propagation, and document-level ACL mapping from source systems into a retrieval layer. You understand why this determines whether governance is real.
• Proficient in Python; comfortable with Go or an equivalent second language.
• Experience with MCP — building servers, not only consuming them.
• Solid cloud-native and systems fundamentals, with Google Cloud Platform strongly preferred (Cloud Run, GKE, Vertex AI, networking, IAM).
• Cost awareness at scale — token and inference spend management across a growing agent estate.

Nice to Have

• Production deployment of agents on Gemini Enterprise / Agent Platform, including custom agents and search experiences.
• Experience with AI evaluation tooling — agent observability platforms, Langfuse, LangSmith, Braintrust, or custom eval frameworks.
• Multi-model routing and fallback across Gemini, Claude, OpenAI or Llama, balancing capability, latency and cost.
• Enterprise data connector work across Google Workspace and Microsoft 365, including entitlement-aware indexing.
• Experience standing up an AI governance function — review boards, architecture checklists, audit evidence — inside a regulated or multi-entity enterprise.
• Containerisation and orchestration (Docker, Kubernetes).
• Fluent use of AI-assisted development tooling to move quickly.