Aether, Almanac, Whitebeard
I worked on multiple data labeling and AI training projects on the Outlier platform, including Aether, Almanac, and Whitebeard. The scope of work involved evaluating AI-generated responses, rating prompt-response quality, checking instruction-following, factual accuracy, relevance, completeness, clarity, tone, safety, and formatting. I also performed response comparison, ranking, rewriting, and quality improvement tasks based on project-specific rubrics. The projects were task-batch based, and I contributed across multiple assigned labeling queues. For quality assurance, I strictly followed the provided annotation guidelines, maintained consistency in ratings, verified factual claims where required, avoided unsupported judgments, protected sensitive information, and incorporated reviewer feedback to improve accuracy and alignment with project standards.