You build and run agentic workflows: systems where models plan, call tools, chain steps and act, not just answer. You understand how agents actually work under the hood, where they go off the rails, and how to keep them on track.
You have built at least one real agentic workflow that did practical work, and you have operated it long enough to know how it fails.
What you will do
Design and build multi-step agentic workflows: tool calling, function calling, planning loops and multi-agent setups.
Wire agents to tools, APIs and data sources, and define what each agent is allowed to do.
Write and tune the prompts, guardrails and state handling that keep an agent reliable.
Operate agents in production: monitor them, catch loops and bad actions, and iterate on failure cases.
Add evals and checks so you know when a change makes the agent better or worse.
What we are looking for
Hands-on experience with agentic stacks is required: LangChain and LangGraph specifically, plus exposure to one or more of CrewAI, AutoGen, LlamaIndex or the OpenAI Agents SDK. You should be comfortable dropping to raw API calls when the framework gets in the way.
One to two years building with LLMs, including at least one real agentic or tool-using workflow shipped in a product.
You know the common agent patterns: function calling, ReAct-style loops, tool use, retrieval, memory and state.
Python or Node, comfortable with async and APIs.
You have seen an agent fail in production and know how you would stop it happening again.