About Mercor

• Build and maintain MCP servers that give agents structured access to Mercor's platform reading project state, taking management actions, and triggering communications.
• Design and implement evaluation frameworks to measure agent decision quality and drive continuous improvement through feedback loops.
• Develop action queues and playbook orchestration systems that allow agents to execute multi-step workflows with conditional logic.
• Instrument agent systems end-to-end — capturing outcomes, closing feedback loops, and enabling hill-climbing on eval metrics.
• Collaborate directly with operators to identify high-leverage automation opportunities and translate them into robust agent behaviors.
• Debug and harden agent systems in production, designing for reliability at scale.
• Contribute to foundational patterns and infrastructure that the rest of the company will build on.
• Strong backend engineering fundamentals in Python (or equivalent modern language).
• Experience building with LLM APIs — including tool use, structured outputs, and multi-step reasoning workflows.
• Comfort designing evaluation frameworks and thinking rigorously about agent failure modes.
• Ability to ship quickly and iterate on systems that interact with the real world.
• Understanding of distributed systems, APIs, and production infrastructure.
• Full-stack experience is a plus — owning features end-to-end will increase your impact.
• Impact: Your work powers how the world's leading AI labs train and test their models.
• Learning: Get early insights into frontier model capabilities months before the market.
• Growth: Work on both infrastructure and research-adjacent projects with fast paths to ownership.
• Generous equity grant vested over 4 years
• A $10K housing bonus (if you live within 0.5 miles of our office)
• A $1.5K monthly stipend for meals
• Free Equinox membership
• Health insurance