Staff Machine Learning Engineer, Agentic Systems

Overview In this role you will help design System 2 for Atlas, the humanoid robot, enabling long-horizon planning and autonomous task execution. You’ll shape the agentic architecture that drives robust autonomy, working with behavior, controls, perception, and product teams to ship production-ready capabilities. You will translate cutting-edge research into practical systems, focusing on safety, reliability, and performance. This is a chance to influence how people interact with humanoids and push the boundaries of robotic autonomy. You will contribute to a strong engineering culture through collaboration, reviews, and documentation. Compensation / Benefitsmedical, dental, vision401(k)paid time offannual bonus structure ResponsibilitiesDesign and implement System 2 architectures for planning, reasoning, memory, and tool-use in humanoid controlDevelop logging, observability and evaluation methods to measure reasoning quality, safety, reliability, and generalizationCollaborate cross-functionally with behavior, controls, perception, and product teams to ship end-to-end capabilitiesEngage with customers and internal stakeholders to align solutions with real-world use casesStay at the forefront of agentic research and translate state-of-the-art techniques into production systemsContribute to engineering culture through code reviews, testing, and thorough system design Key requirements5+ years of professional software engineering, including work on LLM-driven or agentic systemsHands-on experience building or deploying agentic architectures (coding agents, tool-using LLM systems, autonomous task agents)Track record of improving agentic system performance through evaluation and benchmarkingStrong fundamentals in data structures, algorithms, distributed systems, and software architectureDemonstrated ability to ship reliable, maintainable, and well-tested softwareAbility to reason about safety, failure modes, and robustness in autonomous systemscross-functional collaborationeffective communication with stakeholdersproblem solving and critical thinkingLLM-driven/agentic systemscoding agents and tool-use LLM systemssystem design for robotics software