Agent Evaluation and Self-Improvement Engineer

What you’d do

Build evaluation systems for AI agents that catch regressions across harnesses, mine production traces for failures and cost hot spots, and create sandboxed enterprise environments for long end-to-end agent tasks. Full time. Found on the employer's careers page on Oct 17, 2026. Apply through the link.

You’ll need

• Agent evaluation and harness experience
• Mining production traces for failure analysis
• Building sandboxed test environments
• Automated regression testing