AI-Trainee at Yandex (RLHF Evaluator, Data Annotator)
In this role, I evaluated and ranked YandexGPT outputs via RLHF and SFT to assess response quality and factual accuracy in Russian and English. I annotated metadata for standard and agentic AI tasks, including labeling reasoning steps and decision chains. Additionally, I performed validation of citations, fact-checking, QA for language outputs, and wrote rationales for support in reinforcement learning from human feedback. • Evaluated and rated large language model outputs using RLHF and SFT methodologies. • Annotated metadata for both reasoning process and AI agent tasks, labeling logical structures and tool use. • Validated and sourced references, citations, and fact-checks for AI-generated content in both English and Russian. • Conducted multi-faceted QA including grammar, coherence, and technical accuracy.