The role leads Cosmon's core agent intelligence layer, building a system that takes an engineer's intent and executes multi-step workflows across complex desktop engineering software—the hardest and most important problem at the company. They report to the CTO as the technical lead for a team of AI engineers, a user researcher, and domain expert contractors, deeply understanding how mechanical engineers actually work to build an agent that can execute those workflows reliably, correctly, and cost-efficiently. The role owns the full loop from defining capabilities via user story mapping and validation, to building and benchmarking the agent, to setting per-task token budgets and expanding coverage until critical workflows are handled end-to-end, with success measured by task success rate, token efficiency, and coverage. They drive the evaluation framework, translate validated user stories into test cases, and build benchmarks grounded in real user scenarios. They lead architecture decisions for tool-calling, state management across multi-step workflows, error recovery, and model routing, while coaching a team of AI engineers and collaborating cross-functionally with the integration, product, and customer teams in an early-stage, high-stakes environment.