AI Trainer (Contractor)
As an AI Trainer (Contractor) at Techlang AI, I evaluated and ranked LLM-generated responses in Russian and English. I created and refined prompts for language model reasoning and collaborated with a distributed team to calibrate rating criteria. My work focused on ensuring high inter-annotator agreement on sensitive content decisions while providing linguistic expertise for localization quality. • Evaluated LLM responses on helpfulness, harmlessness, and factual accuracy. • Authored, curated, and refined over 2,000 prompts per month in technology, programming, and game development. • Maintained 98% inter-annotator agreement, especially for sensitive and culturally dependent cases. • Automated annotation log conversion to structured JSONL using Python scripting.