Our finance team spends every Monday doing the same manual reconciliation process. Our HR team reviews CVs one by one. Our ops team re-enters data between three different systems. None of this requires human judgment — it requires agents that can read, decide, and act. We need a developer who can build these agents and make them production-ready: not demos, not prototypes, but agents that non-technical teams trust to run unsupervised, with logging, fallbacks, and a clean handoff when things go out of scope.
Responsibilities
Build agentic workflows for our finance, HR, and operations teams
Design tool-use and memory architecture for each agent
Build observability and logging into all agent workflows
Train internal teams on when to trust the agent and when to escalate
Iterate based on user feedback and failure analysis
Requirements
Proven experience shipping agentic workflows in production
Strong Python and familiarity with LangGraph, CrewAI, or AutoGen
Able to design reliable agent architectures with human-in-the-loop fallbacks
Experience integrating external tools and APIs into agent loops
Good communicator — works closely with non-technical operations teams