AI Trainer, RLHF Annotator & Reasoning Evaluator (Freelance / AI Training Platforms)
Text
RLHFPerformed large-scale RLHF annotation by reviewing and rating 5,000+ AI-generated responses for accuracy, reasoning quality, helpfulness, and safety using structured multi-criteria rubrics. Conducted comparative preference ranking and side-by-side evaluation of model outputs to produce RLHF training signals that informed reward model improvement cycles. Built and applied adversarial and boundary-testing prompt strategies, then documented structured feedback for evaluation leads to improve model logic. • Preference ranking and comparative rating of model outputs for RLHF training • Correctness and quality evaluation using structured scoring rubrics and QA checks • Adversarial/boundary prompt testing to surface reasoning and safety gaps • Dataset labeling and curation with consistent metadata tagging and standards