Software Engineer III, AI Agent Builder — FinOps & Technology Cost Value Management

Who We Are Wayfair’s FinOps and Technology Cost Value Management (TCVM) team is building the next generation of autonomous engineering workflows. We treat cloud optimization and AI cost management not as passive reporting tasks, but as active software engineering challenges. We are creating "agentic digital employees"—intelligent, stateful AI agents and graph execution loops that operate within engineer\-native tools like Cursor and Devin. By connecting LLM\-driven agents to Google Cloud Platform (GCP) advisories, codebase graphs, and cost attribution pipelines via Model Context Protocol (MCP) and Agent\-to\-Agent (A2A) communications, we automate cost\-saving infrastructure changes safely at scale. What You’ll Do \* **Agentic Architecture \& Graph/Loop Engineering:** Design, build, and maintain stateful agentic workflows, execution loops, and multi\-agent topologies (using state graphs, DAG execution engines, and cyclic recovery loops). Implement Model Context Protocol (MCP) servers and Agent\-to\-Agent (A2A/ACP) communication interfaces to connect AI developer agents (Devin, Cursor, GitHub Copilot) directly to internal databases, CI/CD pipelines, and FinOps platforms. Engineer agentic loops capable of multi\-step reasoning, dynamic tool selection, error detection, and self\-correction when parsing code or cloud configurations. \* **Autonomous FinOps \& IaC Pull Request Generation:** Build agents that consume GCP spend advisories (GKE, BigQuery, Compute Engine) and automatically generate context\-aware Infrastructure\-as\-Code (Terraform, Helm) pull requests. Embed application dependency checks and performance guardrails into agent\-generated code proposals to ensure changes do not violate service SLOs. Develop automated evaluation datasets (evals) to measure recommendation accuracy, PR quality, and execution safety before human review. \* **AI Workload Economics \& Spend Management:** Build tracking and attribution tooling to measure AI consumption across Wayfair, including token usage, cost\-per\-API\-call, model inference latency, and vendor model selection costs. Engineer real\-time anomaly detection workflows and automated triage loops that flag sudden spikes in cloud or AI/SaaS expenditures to responsible team owners. Construct automated showback/chargeback telemetry models to give engineering domains granular visibility into their AI footprint. \* **Human\-in\-the\-Loop (HITL) Guardrails \& Security Governance:** Design robust policy engines, permission boundaries, and step\-by\-step approval gates to ensure AI agents execute changes with appropriate access controls. Embed safety checks into CI/CD pipelines, preventing unauthorized schema migrations, dynamic code execution flaws, or unreviewed infrastructure modifications. Standardize reusable agent building blocks, SDKs, and templates across Wayfair's technology organization. **We Are a Match Because You Have:** \* 5\+ years of professional software engineering experience building production services, microservices, or platform tools. \* High proficiency in Python or TypeScript/Node.js, with experience in Go or Java as a plus. \* Proven experience constructing AI agents, tool\-calling workflows, state graph execution (e.g., LangGraph, AutoGen, or custom state machines), and prompt/eval engineering frameworks. \* Deep understanding of MCP (Model Context Protocol), A2A patterns, REST/gRPC APIs, and orchestration of developer tools (Devin, Cursor, GitHub Copilot). \* Hands\-on experience with Google Cloud Platform (GCP), GKE (Kubernetes), BigQuery, and Infrastructure as Code (Terraform, Helm). \* Familiarity with cloud spend tracking, token economics, API cost attribution, or a FinOps.org Certification (Practitioner/Engineer). \* Experience with unit/integration testing for LLM outputs, CI/CD pipeline development, and secure enterprise authentication/authorization (OAuth2, IAM, RBAC).