AI Training Generalist
As an AI Training Generalist at Mercor, I delivered high-precision feedback on model responses in an RLHF framework. My work involved testing response pairs, auditing prompts, and ensuring safety, factuality, and guideline compliance. I also reviewed annotations and directly labeled sentiment for algorithmic safety evaluation. • Conducted model response pairwise comparisons with over 99% accuracy. • Audited prompts and submissions in alignment with safety and formatting rules. • Labeled over 30 sentiment labels per hour for compliance verification. • Provided constructive feedback to junior annotators.