Founding Harness Engineer — AI Agents (Contract-to-hire)

Own the runtime, prompts, eval and regression harness, and guardrails for an early-stage 'AI coworker' agent that lives in a team's everyday tools.

The partner startup's agent works inside a team's chat and productivity tools. The open seat owns how it behaves: what context it pulls, which tool it picks, when it asks permission, and how it survives long tasks with retries and flaky APIs.

It starts as a contract with a path to full-time. Prompt changes are treated like code changes, so the engineer owns the eval and regression harness as well as approval gates and tenant data separation.

What you would own

• Build the agent runtime: context loading, tool choice, approval waits, and when to act or stay quiet.
• Write and keep tuning system prompts, tested like code.
• Own the eval and regression harness for agent behavior.
• Make long-running work survive interruptions and retries.
• Add guardrails: approval gates for risky actions and strict customer data separation.

What they're looking for

• You have shipped tool-using LLM agents to real users in production.
• Hands-on experience building eval and regression suites for LLM behavior.
• Understanding of safe tool execution, prompt injection, and multi-tenant data risk.
• Solid backend skills in Python or TypeScript.
• You must be based in the US.

Location: United States (Pacific-time overlap preferred). Hours: Contract to start · path to full-time. Pay: Rate not listed.

Bolder Apps is hiring on behalf of an unnamed early-stage partner startup building an AI coworker agent.