Agent Harness Engineer

This is a fully remote position, open to applicants in Germany.

📋 Description

• Develop the agent runtime: the cycle where it plans, acts, and verifies its own work rather than claiming success it hasn't earned.

• Ensure the agent is proficient with tools: integration composition, memory recall, and skill files that provide the necessary functions.

• Expand the agent's capabilities: reliable memory, self-generated skills, and autonomous task execution.

• Enhance measurable performance: evaluations, identifying regressions, and transforming every setback into a solution.

• Deliver products: features that are accessible to users via Slack and Teams within hours.

• Address whatever needs to be constructed. Small team, extensive scope.

⛳️ Requirements

• You have personally developed or significantly altered an AI agent harness and can articulate the trade-offs involved.

• You have successfully implemented an agent that reliably closes the loop: through tests, linters, type checkers, and browser verification, ensuring it does not mistakenly report success.

• You have created custom skills, commands, CLIs, or MCP servers to expedite your own agent-based coding.

• Agentic coding is part of your everyday workflow, and you have rigorously tested models and harnesses on the cutting edge.

• You possess systems thinking. You navigate complex technical trade-offs and design with an understanding of potential failures at scale.

• You prioritize speed, meeting high standards. You deliver today, not on Thursday.

• Versatility is key. You can transition from agent runtime to React component to deployment script within a single afternoon.

• You have a genuine curiosity about how AI functions. You have read relevant research papers and formed opinions about evaluations.

🏝️ Benefits

• Competitive salary at the forefront of the market.

• A level of ownership that is unique to this stage of development.