Member of Technical Staff, Agents

ABOUT THE ROLE Join a small, senior team building autonomous AI assistants for executive workflows. You will develop agent capabilities and evaluation systems that help ensure real-world tasks are completed reliably, with measurable quality. WHAT YOU'LL DO - Build agents that complete delegated work end to end for real users. - Define reliable measures of success for agent tasks and maintain representative evaluations. - Develop new agent capabilities and validate them before they reach users. - Investigate production failures and deliver fixes that address underlying causes. - Build systems to detect known failure patterns and emerging issues at scale. - Improve agent recovery when tools or websites fail. - Create repeatable methods for engineers to measure the impact of agent changes. WHAT WE'RE LOOKING FOR - At least 3 years of data science or machine learning experience building evaluation systems, metrics, or quality measurement for products. - Production-quality Python and SQL skills, including experience analyzing large datasets and building automated pipelines. - Experience designing evaluation frameworks, datasets, measurable criteria, and appropriate metrics. - A track record delivering production AI, machine learning, agent, or agent-evaluation systems for real users. - Experience improving agent reliability, owning systems end to end, and measuring results in production. - Familiarity with statistics, experiment design, and realistic end-to-end simulations; experience with fine-tuning, post-training, or reinforcement learning is valuable. - Experience with Python, Django, React, and TypeScript; familiarity with evaluation tools such as Braintrust or Raindrop is useful. - Demonstrated progression and sustained experience in a role, with the ability to work through ambiguity and ship quickly. COMPENSATION & BENEFITS Base salary is $200,000 to $300,000 USD annually. Visa sponsorship is not available. LOCATION On-site in Palo Alto, California, with four days a week in the office.