Mobarak Farzad

I've evaluated AI-generated responses and training data for Turing, Prolific, Scale AI, and Alignerr, assessing accuracy, relevance, reasoning, coherence, instruction-following, and safety. I create prompts and examples, rank model outputs, identify hallucinations and bias, and provide structured feedback that improves model quality.

At CACTUS Ltd, I supervised 15+ remote translators while reviewing English and Dari content and cleaning multilingual NLP datasets. I bring native-level Dari/Farsi, C2 English, and working Pashto, Hindi, and Spanish to multilingual AI evaluation, translation quality assurance, and cross-cultural language work.