Senior AI Image Evaluator & Data Specialist
I served as a specialist evaluator for high-priority RLHF projects, specifically focusing on the Omni Elo and Outlier platforms. My core responsibility involved the rigorous assessment of text-to-image and image-to-image model outputs. I utilized a specialized three-step evaluation protocol that prioritized strict instruction following and prompt adherence as the primary metrics for success, ensuring that models were penalized for missing constraints regardless of aesthetic quality. The scope of my work included identifying subtle model hallucinations, assessing visual realism, and spotting "AI-generated" artifacts. I provided high-volume, evidence-based feedback in a structured format to help developers refine model alignment. By maintaining a "logic-first" approach to ranking, I consistently met high quality-control standards, contributing to the improved accuracy and safety of cutting-edge generative systems.