AI Data Trainer & RLHF Annotator | Freelance Contractor (AI Platforms)
As an AI Data Trainer and RLHF Annotator, I evaluated AI outputs by providing expert-level preference ratings and detailed logical justifications. I authored comprehensive rationales to clarify model instruction-following, truthfulness, safety, and tone boundaries. I generated 'Gold Standard' rewrites for incorrect responses to be used in Supervised Fine-Tuning datasets. • Processed over 90 complex reasoning cases in single 12-hour work sprints with zero rejection. • Rated outputs across multi-dimensional scales, distinguishing nuanced failures such as hallucinations and stylistic errors. • Flagged critical model failures in Portuguese-English translation and ensured model guidance improvements. • Created direct training data to fine-tune model completions within a structured workflow.