AI Trainer / AI Data Annotator & Evaluator
As an AI Trainer and Data Annotator & Evaluator at Outlier AI, I evaluated outputs from large language models such as GPT-4 and Claude. My role included annotating and labeling AI-generated responses using structured rubrics and providing feedback that informed model improvement. I contributed to adversarial prompt design and iterative RLHF workflows to enhance model reasoning, factual accuracy, and safety. • Reviewed and ranked LLM outputs for coherence, helpfulness, and alignment with user intent. • Designed and implemented adversarial prompts to test model performance across technical and general domains. • Applied domain expertise to identify hallucinations, bias, and inconsistencies in LLM responses. • Collaborated to set annotation standards and contributed to AI safety research documentation.