AI Engineer – Agentic Development for Seattle, WA

EXL is hiring two AI Engineers to join a strategic agentic AI engagement with a leading specialty retailer. This is a hands-on build role embedded with the client's engineering team, focused on shipping production agentic systems that automate real business workflows.

The role requires designing, building, and evaluating multi-agent systems — from agent orchestration and tool design through evaluation and context management. The right person is comfortable owning the full lifecycle of an agent's development: from design through production deployment, working from requirements the client has already scoped.

This is a short-term, high-visibility engagement with potential to extend or convert as the client's AI program grows.

• Design and build production agents and multi-agent orchestration systems — not prototypes or one-off experiments.
• Apply sound judgment on agentic design trade-offs: orchestration patterns, multi-agent coordination, tool design, and MCP (Model Context Protocol) integration considerations.
• Build and run programmatic evaluations for agents: metrics, offline/online evals, and failure-mode analysis.
• Apply context engineering practices: context assembly, memory/state management, and context handling under token, latency, and cost constraints.
• Ship production-ready code through the full SDLC, working largely independently from requirements the client's team has already defined.
• Partner with client engineers and stakeholders; stay cognizant of end-customer and user experience even when requirements are handed off rather than gathered directly.
• Contribute to engineering standards and help raise the broader team's working fluency with generative AI by example.

• Software Engineering: 5+ years of experience with strong fundamentals — production-ready code, full SDLC ownership, and real deployment experience, not just development.
• Generative AI / Agentic Systems: Recent (within the last year), hands-on experience building and shipping agentic systems — open source or production — beyond experimentation.
• Foundation Models: Working knowledge of modern foundation models and current agentic design approaches and trade-offs.
• Evaluation: Practical experience with agent evaluation — programmatic metrics, offline/online evaluation, and failure-mode analysis.
• Context Engineering: Practical experience with retrieval, context assembly, and memory/state management under real constraints.
• Frameworks: Comfortable working framework-agnostic. Exposure to any of LangGraph, AutoGen, CrewAI, Semantic Kernel, Claude Agent SDK, or the OpenAI Assistants API valued as evidence of transferable concepts, not a checklist requirement.
• Communication: Strong written and verbal communication; comfortable operating from client-defined requirements with minimal hand-holding.

• Hands-on experience specifically with the Claude Agent SDK.
• Experience with LLM cost management and optimization.
• Strong stakeholder communication and requirements-translation skills.
• Domain experience in Retail or E-commerce — customer journey data, transaction analytics, or similar.