Evaluate autonomous AI agents with Outlier using technical feedback, agent debugging, and complex systems integration. This freelance work is remote.
Outlier helps creative companies improve AI agents by collecting human feedback. As an engineer, you will help train large language models to operate as proactive, multi-step systems. Projects involve architectural workflows, live tools, and system behaviors requiring careful analysis.
Key Responsibilities: Debug agent traces and investigate system behavior
Write clear evaluation rubrics for agent outputs
Stress test agents against complex interactions
Identify privacy leaks and authority escalation failures
Review indirect prompt injection risks
Requirements: At least two years of relevant professional experience
Backend engineering, AI automation, or integration background
Maintain production-grade modular software systems
Handle persistent state and multi-turn interactions
Provide precise technical feedback on complex behaviors
Use agent evaluation, debugging, and security analysis skills to help Outlier improve autonomous AI systems like OpenClaw.