AI Trainer / Generalist Rater (Remote, Freelance) — Outlier & Multimango
Served as an AI Trainer and Generalist Rater evaluating AI-generated responses for accuracy, clarity, and helpfulness across a range of generalist tasks. Compared and ranked competing model outputs to support reinforcement learning from human feedback (RLHF) and improve response quality. Wrote and refined prompts to test model behavior while fact-checking outputs and flagging hallucinations, bias, and unsafe or incorrect content.• Evaluated response quality across multiple task types (accuracy, clarity, helpfulness).• Ranked competing outputs for RLHF-style preference signals.• Authored and iterated prompts to elicit better model behavior.• Performed claim verification and safety/bias checks on model responses.