Multi-Agent Orchestration System (LangGraph Supervisor)

Description: We're building a system that requires several specialized AI agents working together under a supervisor, and we need an engineer who has actually built multi-agent orchestration before, not just single agents or simple chatbots. The hard part here isn't any one agent, it's the coordination: routing work to the right agent, passing state cleanly between steps, and keeping the whole thing reliable in production. What we're building: We have a workflow that needs to be broken across multiple specialized agents: A research agent that gathers and structures information A drafting agent that produces the output A validation agent that checks the draft against requirements A supervisor that decides which agent runs when, routes work between them, and manages the overall state Responsibilities: Design the supervisor architecture and the agent graph in LangGraph Build 3 specialized sub-agents (research, drafting, validation) with clear responsibilities Implement state passing between agents so context isn't lost across steps Add retries and error handling so a single failure doesn't break the whole run Add a human-in-the-loop approval gate before any final output is committed Keep behavior predictable and debuggable in production Requirements: Real, demonstrable experience with multi-agent orchestration (LangGraph specifically preferred) Strong with LangChain, LangGraph, and Python Understanding of supervisor/sub-agent patterns, state management, and conditional routing Experience making agent systems reliable, not just working in a demo Clean, well-documented, maintainable code Bonus: Experience with human-in-the-loop approval flows LangSmith or similar for tracing/observability of agent behavior