Nzube Ndibueze

Reviewed short video and audio clips to determine whether written statements about them were accurate (Yes/No/Not sure), providing an evidence-based reason for each decision. Earned a promotion to reviewer after achieving a 90% quality score through hidden checks. As a reviewer, audited other annotators’ work, approved or returned it with concise corrective feedback. Improved throughput from ~47 to 250+ tasks per hour while maintaining accuracy, completing a few thousand tasks. Raised guideline conflicts with examples rather than guessing to ensure consistent interpretation.

Evaluated social media posts, videos, images, and linked articles for misinformation, scams, and misleading claims. Identified central claims, assessed verifiability, and researched using credible sources (e.g., health authorities, regulators, and fact-checkers). Rated claims as true, false, partly false, or unverified, and flagged scam patterns, impersonation accounts, and manipulated or out-of-context media. Applied policy guidelines consistently while meeting productivity and quality targets.

Evaluated and refined multi-turn dialogue outputs against rigorous quality guidelines, improving semantic alignment, tone, and factual consistency. Performed RLHF workflows, fine-grained text annotation, and prompt-response ranking to improve AI response accuracy. Proofread and copyedited English model outputs to reduce grammatical ambiguity and hallucinations. Participated in calibration sessions to adapt to shifting taxonomies and compliance criteria.