Design and build long-running agents that own multi-step growth and algo workflows, from feature research pipelines to creative generation and validation.
Build the feedback loops that make agents trustworthy: evals, assertions, state management, self-correction, and human-in-the-loop review for both technical and non-technical users.
Own AI cost reduction across the company (Claude / Cursor and our products), including monitoring token spend, model routing, context optimization, and turning repeated workflows into cheap reusable skills.
Own the company's AI knowledge hub, including how knowledge is routed, kept fresh, and made self-improving from real agent sessions.
Define the metrics and guardrails that let agents run for hours or days unattended without silent degradation or runaway cost.
Partner with every department to identify workflows worth automating, and prove the automation actually works.
Continuously evaluate new models, tools, and agent frameworks, and decide what is worth adopting based on real results.
Work in a results-driven, high-paced, rewarding environment.
4+ years of hands-on experience as a Software Engineer, Algorithm Developer, or similar, with proven ability to ship robust, production-grade systems.
Strong problem-definition skills: you turn "make it better" into a metric, a baseline, and a decision rule before writing code.
Algorithmic thinking and comfort reasoning about optimization, evaluation, and trade-offs.
Strong Python, and comfort building and debugging systems with non-deterministic components.
Hands-on experience building with LLMs beyond chat: agents, tool use, evaluation harnesses, or automated pipelines.
Strong communication across disciplines. Able to talk shop with engineers, data scientists, creative, and growth alike.
Experience with cloud environments (GCP preferred) and data tooling (SQL / BigQuery a plus).
A cost-aware mindset: you treat tokens and dollars like memory and CPU.
Strong analytical mind with proven problem-solving abilities. Independent, proactive, and able to drive ambiguous projects to completion.
