Description
Summary We're looking for an experienced AI engineer to build a customer-support AI agent that can actually take actions, not just answer questions. Most "AI chatbots" just retrieve answers — we need an agent that calls our internal tools, decides what to do, and knows when to hand off to a human. What the agent needs to do: * Call our internal tools to look up orders, check status, and create support tickets * Decide, based on the conversation, which tool to use and when * Escalate to a human agent when the issue is outside its scope or high-stakes * Behave predictably — no hallucinated actions, no acting on bad data Responsibilities: * Build the agent using LangChain with a tool-calling architecture * Wire up 3-4 tool integrations to our internal APIs (order lookup, ticket creation, status checks) * Add guardrails and validation so the agent only takes safe, intended actions * Implement a clean human-escalation path for cases it shouldn't handle alone * Ensure reliable, predictable behavior in production Requirements: * Real experience building tool-calling AI agents (not just chatbots) * Strong with LangChain and Python * Experience integrating agents with internal tools/APIs * Understanding of agent guardrails, validation, and reliability * Clean, well-documented code Bonus: * Experience with human-in-the-loop escalation flows * Prior customer-support automation work If you've built tool-calling agents before, please share a relevant example or portfolio link.