Staff AI Engineer, Consumer Agents

Role overview Lead the AI engineering behind consumer agents that carry out multi-step tasks on a person’s behalf. You’ll develop the planning, negotiation, and verification capabilities that help agents work across services, use tools over multiple turns, and handle failure responsibly.

Responsibilities - Architect the agent loop, including planning, sub-agent dispatch, tool selection, memory, and graceful failure. - Build an evaluation system covering task completion, factual accuracy, refusals, and latency limits. - Choose models and manage prompt routing across external and internal providers, balancing cost and quality. - Extend agent capabilities to handle tasks such as negotiating refunds, changing providers, and closing accounts. - Mentor engineers on responsible agent design.

Requirements - At least seven years of engineering experience, including two or more years building production LLM systems. - Demonstrated experience shipping multi-step agents beyond prototype or demo projects. - Strong understanding of evaluation methods, including LLM-as-judge, golden-set sampling, and regression tracking. - Working fluency in Python and TypeScript. - Ability and interest to work fully remotely.

Nice to have - Experience with agent products where errors can have financial consequences. - Background in reinforcement learning, planning, or program synthesis. - Published or open-source work on agent infrastructure.