Four or more years of experience with tool-using LLM agents in production environments.
Hands-on experience in system prompt design with a focus on iterative improvement.
Proven ability to build evaluation and regression test harnesses for LLM or agent behavior.
Practical knowledge of safe tool execution and approval gates.
Experience managing multi-user or multi-tenant data separation.
Demonstrated ownership of technical projects from prototype to production.
Responsibilities
Design and build the agent runtime for context gathering and decision-making.
Write and refine system prompts to promote useful initiative while minimizing risks.
Create evaluation and regression tests for real tasks and tool usage.
Enhance reliability for long-running tasks and manage interruptions effectively.
Implement safe execution boundaries with approval gates and data separation.
Collaborate with leadership and early users to shape product direction.
Benefits
Founding equity opportunities.
In-person collaboration in a dynamic San Francisco environment.
Full Job Description
Build the runtime and evaluation systems behind an AI coworker for teams. In this founding role, you will help make tool-using agents proactive when useful, careful when approval is needed, and dependable in real work. What You'll Do
Design and build the agent runtime, including how it gathers context, decides when to respond, selects tools, requests approval, and completes work.
Write and refine system prompts that encourage useful initiative without noisy or unsafe behavior.
Create evaluation and regression tests for real tasks, tool use, and failure cases, and use them to catch regressions from prompt changes.
Improve reliability across long-running tasks, interruptions, and retries.
Implement safe execution boundaries, including approval gates and separation of multi-tenant data.
Work directly with company leadership and early users to shape the product.
What We're Looking For
Four or more years of relevant experience, including production deployment of tool-using LLM agents, rather than demos or thin prompt wrappers.
Hands-on system prompt design, iterated and improved through testing.
Experience building evaluation and regression test harnesses for LLM or agent behavior.
Practical experience with safe tool execution, approval gates, and multi-user or multi-tenant data separation.
Strong ownership across broad technical problems, from prototype through production, and the ability to explain tradeoffs.
Interest in a founding role and in-person collaboration in San Francisco.
Compensation & Benefits
Base salary up to $200,000, plus founding equity. Equity percentage and vesting details are to be confirmed. Location
On-site in San Francisco, United States. Candidates should be in the Bay Area or willing to relocate.