Design and ship A/B tests across agent-facing surfaces including MCPs, CLIs, skills, APIs, and documentation.
Analyze behavioral and experiment data and turn findings into shipped product improvements.
Form and test production hypotheses about how agents discover, interpret, and select tools and capabilities.
Track emerging agent tools, models, and harnesses, including Hermes and OpenClaw.
Own the agent-experience problem end to end and operate independently in an emerging space.
Requirements
Demonstrated ability to ship high-quality products quickly rather than prototypes or presentations.
Experience designing A/B tests, interpreting behavioral data, and acting on results.
Familiarity with agent-facing surfaces such as MCPs, CLIs, skills, APIs, and documentation.
Active interest in emerging AI tools, models, and harnesses, including Hermes and OpenClaw.
Genuine curiosity about how agents select tools, interpret capabilities, and interact with products.
Strong independent ownership and high agency in an environment without a predefined playbook.
Experience as an early startup engineer, former founder, prolific open-source contributor, or someone with impressive public work.
At least 5 years of experience and legal authorization to work in the United States; visa sponsorship is not currently available.
Benefits
$235,000–$260,000 annual salary and competitive equity.
15 days of mandatory PTO, with additional time available by request, plus 12 weeks of fully paid parental leave.
$100 monthly wellness stipend and up to $1,000 annually for learning and development.
Team offsites and a three-month paid sabbatical after four years.
U.S.-based full-time employees receive medical, dental, and vision coverage, employer-paid life and disability insurance, optional supplemental insurance, telehealth, a 401(k) plan, FSAs, commuter benefits, and pet insurance.
San Francisco employees receive office snacks, drinks, team lunches, and a loaner electric bike.
The role is full-time and based in San Francisco, California.
The hiring process includes a paid 1–2 week work trial at a contractor rate.