AI Data Contributor | Toloka Platform
I evaluated and ranked AI-generated responses for helpfulness, accuracy, and safety to improve large language model outputs. I created prompts and ideal responses for AI conversational training and identified logical errors and hallucinations across various topics. Throughout my tenure, I consistently maintained a quality score above 95% over 500+ completed tasks. • Performed RLHF tasks on text-based conversational AI. • Assessed and scored responses based on detailed guidelines. • Authored original prompts for fine-tuning dialogue models. • Contributed to safety evaluations and error flagging.