Sr. Full Stack Software Engineer – AI Agents

About the role

A contractor program is looking for a senior full stack software engineer to build applications around multi-agent systems developed with Google's Agent Development Kit (ADK) on Google Cloud Platform. The engineer builds the agents, the backend services and APIs behind them, and the web interface used by users. The role suits someone who has built agents with ADK, deployed them to production on GCP and kept them running with evaluation and feedback loops. AI coding tools are used daily, with several agents running at once.

Key responsibilities

• Design the multi-agent architecture, including agent roles, delegation, tool boundaries and state handling.
• Implement agents and tools with ADK, integrating with customer and third-party systems via REST and MCP.
• Build the web frontend and backend services that expose the agents, including user workflows and the APIs they run against.
• Deploy agents and services to GCP on Cloud Run and Agent Engine, with CI/CD, managed environments and releases.
• Build evaluation suites and feedback loops that measure agent output, surface regressions and drive changes to prompts, tools and agent structure.
• Instrument agents and services with tracing and logging for production debugging.
• Use Claude and multi-agent workflows to speed up design, implementation, test generation and code review, while keeping accountability for the correctness of all delivered code.
• Deliver unit, integration, end-to-end and agent evaluation test coverage, take part in code reviews and document the architecture and operating procedures.

Required profile

• Professional full stack software engineering experience, including production delivery on Google Cloud Platform.
• Bachelor's degree in Computer Science or equivalent practical experience.
• Experience building and deploying AI agents with ADK, including evaluation sets, automated scoring, tracing and revising prompts and tools.

Required skills

• ADK agent development, including agent definitions, tool design and session and state management.
• Multi-agent system design: delegation, sequential, parallel and loop workflows, and clear agent boundaries.
• Frontend with TypeScript and a modern component framework, streaming responses and authenticated sessions.
• Backend services and APIs, relational database design (e.g. PostgreSQL) with versioned schema migrations.
• Authentication and authorization with OAuth 2.0 / OIDC and JWT.
• Deployment on GCP using Cloud Run and Agent Engine; Gemini models on Vertex AI, prompt design and function calling.
• Tool integration over MCP and REST APIs, and daily use of Claude and other AI coding assistants.
• Automated testing and CI/CD across the stack, including frontend and end-to-end tests.