AI Trainer / AI Evaluator – Outlier (Evaluation, ranking, and safety review)
Evaluates AI-generated text and image responses for quality, accuracy, and relevance to the given prompts. Performs response ranking, fact-checking, safety reviews, and prompt assessment to score model outputs consistently. Flags factual inaccuracies, instruction-following issues, and image-region errors to support model improvement. • Ranked multiple responses based on quality criteria • Conducted factual verification and accuracy checks • Performed safety and prompt assessment reviews • Identified errors in instructions and image regions