AI Trainer & Data Annotator
As an AI Trainer & Data Annotator at Mindrift (Toloka AI), I evaluated and rated LLM-generated responses to enhance language model capabilities. I conducted adversarial red-teaming, authored gold standard responses, and engaged in complex prompt and response writing. My work focused on both technical tasks and language-based evaluation. • Evaluated AI model output for helpfulness, honesty, and harmlessness (HHH) • Authored gold-standard reference responses for prompt completion tasks • Conducted red-teaming for prompt injection and vulnerability identification • Classified and reviewed code snippets for correctness and bug identification