Join Emergent, a leader in autonomous coding agents, as an AI Agent Architect. We're revolutionizing software development by generating, testing, and deploying production applications directly from plain-language intent. Our platform powers millions of real applications globally, backed by top-tier investors and a team of industry leaders.

This role is for builders who thrive on ownership, speed, and making an impact at a global scale. You'll be instrumental in shaping the future of AI-driven software creation by focusing on correctness, reliability, security, and scale in real-world production systems.

As an AI Agent Architect, you will transform raw LLM and system capabilities into tangible improvements in agent performance. You'll own the end-to-end development loop, from ideation and experimentation to validation and deployment. This multifaceted role sits at the nexus of product, engineering, and applied research, empowering you to define what makes our agents better, how we measure success, and what ultimately goes live at scale.

Key responsibilities include: - Developing deep insights into agent behavior, identifying success patterns and failure modes. - Analyzing production data to uncover subtle regressions and bottlenecks. - Designing and executing high-impact experiments to enhance agent quality and reliability. - Building robust evaluation frameworks to measure agent performance at scale. - Ensuring rigorous deployment through clear metrics, evaluation gates, and rollback strategies. - Tackling complex challenges in context engineering, memory, tool use, and long-horizon execution. - Making critical decisions in probabilistic systems regarding regressions, wins, and promising research directions.

We are seeking an experienced professional with 5+ years of hands-on software development and end-to-end ownership experience. A minimum of 1 year working with agentic systems and their evaluation is essential. You should possess strong systems thinking capabilities combined with sharp product judgment and fluency in languages like Python or SQL for analysis, experimentation, and debugging.

Comfort with reasoning about noise, confounds, distribution shifts, and selection effects is crucial, as is an intrinsic motivation for sifting through large volumes of agent behavior to identify rare failure modes. A deep, current interest in LLMs, agents, and emerging model capabilities is expected. You should be an independent operator with leadership presence, capable of scoping your own work and advocating for sound ideas while maintaining a bias toward velocity without compromising measurement integrity.