Gemini AI architect San Ramon CA Local only

HMG America LLC is an IT consulting and services company specializing in software and web development, staff augmentation, and professional services. One of our direct clients is seeking a Gemini AI Architect in San Ramon, CA for a hybrid role requiring 3 days onsite. Preference will be given to local candidates who can attend in-person meetings as needed.

About the Team

We are standardizing our agentic capabilities on Google Gemini Enterprise as the foundational platform, building custom agents, tools, and governed workflows on top of the Agentic Studio Platform. These are connected to enterprise data and applications through MCP (Model Context Protocol). This is a high-impact, high-autonomy team where you'll collaborate with engineering, product, and operations teams to bring AI capabilities to life.

About The Role

As an Applied AI Developer on the Enterprise AI team, you will evaluate AI applications, build tools and platforms, and enable teams to adopt modern AI development patterns. These include LLM orchestration, agentic workflows, and model governance on Google Gemini Enterprise. You will design and ship Custom Agents using the Agent Development Kit (ADK), register them in the Agent & Tool Registry, and connect them to enterprise systems via BYO and vendor-managed MCP servers. You will enforce Agent Identity, Security, and Observability end-to-end while staying hands-on and outcomes-oriented.

Our AI Platform Environment

• Platform surface (gemini.google.com): Google Enterprise business apps, Custom Agents, and enterprise Search UX as primary experiences for internal users.
• Agentic Studio Platform: ADK & runtime, model training, Agent & Tool Registry, Agent Identity, Agent Security, and Agent Observability for building, governing, and operating agents.
• Multi-model / Bring Your Own LLM (BYOLLM): Flexible model routing across Gemini, Claude, LLAMA, and OpenAI, selecting the right model per use case, cost, and quality target.
• Integration & data layer: Enterprise data sources (Google Workspace, Microsoft 365, and others), BYO MCP servers (application/tool-centric), and vendor-managed MCP servers & agents.

Responsibilities

• Assess where agentic approaches outperform conventional solutions and own the quality bar by building automated evals, simulation tests, and regression frameworks integrated with Agent Observability.
• Design agentic systems on the Agentic Studio Platform, including Custom Agent development with ADK, tool orchestration, agent reasoning, memory, and MCP integrations (both BYO and vendor-managed servers).
• Define and implement AI governance patterns, including guardrails, data lineage, auditability, and responsible AI practices using Agent Identity, Agent Security, and the Agent & Tool Registry.
• Enable multi-model flexibility (BYOLLM) by implementing model routing and fallback across Gemini, Claude, LLAMA, and OpenAI, balancing capability, latency, and inference cost.
• Drive adoption through pilots, proofs-of-concept, and scalable implementations across engineering teams, delivered through gemini.google.com surfaces (business apps, Custom Agents, and Search UX).
• Collaborate with business functions, product, security, and platform teams to translate AI use cases into production-grade, end-to-end solutions connected to enterprise data sources across Google and Microsoft ecosystems.

Requirements

• 7+ years of professional software engineering, with at least 2 years focused on applied AI in production systems.
• Proficient in Python and/or Go; comfortable reading and writing in the other.
• Proven experience building and scaling multi-agent or agent-driven systems in production, including real-world operational ownership.
• Hands-on experience with Google Gemini Enterprise and the Agent Development Kit (ADK), or comparable enterprise agent platforms, including agent runtime, agent/tool registration, identity, and observability.
• Hands-on experience with modern agent ecosystems, including frameworks (e.g., Google ADK, LangGraph, Mastra, Claude Agent SDK), observability and evals tooling (e.g., Agent Observability, Langfuse, LangSmith, Braintrust), MCP implementations, and leading AI SDKs across a multi-model/BYOLLM environment (e.g., Gemini/Vertex AI, Anthropic (Claude), OpenAI, LLAMA).
• Strong systems and backend architecture fundamentals, including designing scalable, reliable systems and handling infrastructure, performance, failure modes, cost, and deployment concerns.
• Good understanding of cloud-native environments, with Google Cloud (GCP) and Vertex AI strongly preferred (and/or AWS), including compute, storage, networking, and managed AI services.
• Experience designing and integrating with enterprise APIs (REST, GraphQL), including authentication and authorization patterns (OAuth2, SAML, API keys, RBAC), and connecting agents to enterprise data sources across Google Workspace and Microsoft 365.
• Comfortable working with backend databases (SQL and NoSQL), writing queries, understanding data models, and building data access layers that enforce role-based access control aligned with Agent Identity and Agent Security.
• Strong cross-functional collaborator and communicator, able to partner with Product, Operations, and domain experts to deliver end-to-end systems with measurable real-world impact.
• A force-multiplier on the team, raising the bar for clarity of thinking, system design standards, and team execution.

Nice to Have

• Direct experience deploying agents on Google Gemini Enterprise / Agentic Studio Platform (Custom Agents, Search UX, business apps) in production.
• Experience with AI evaluation tooling (Agent Observability, Langfuse, LangSmith, Braintrust, or custom eval frameworks).
• Experience building custom MCP servers (both application/tool-centric BYO servers and vendor-managed servers & agents).
• Familiarity with containerization and orchestration (Docker, Kubernetes) and GCP-native deployment (Cloud Run, GKE).
• AI-native builder with high velocity and ownership: intellectual curiosity, rapid adoption of new tools, bias to action, and the ability to drive ambiguous problems from concept to production.
• Hands-on experience with inference cost optimization across a multi-model/BYOLLM setup, managing spend as agent deployments scale.
• Experience using AI-powered coding agents (e.g., Claude Code, GitHub Copilot, Cursor, Windsurf) to accelerate development workflows, including rapid prototyping, code generation, debugging, and test writing.
• Experience with RAG (Retrieval-Augmented Generation) architectures and document retrieval pipelines, including vector databases, embedding models, chunking strategies, and hybrid search for building agents grounded in enterprise documentation and data sources.

Duration: 6 months or longer.