AI Safety LLM Trainer (Korean C1+ English Required)

Project Overview / Instructions

Candidates must have a bachelor’s degree or higher in a relevant field; be near-native or native in Korean; have at least C1 English proficiency; and show experience in Trust & Safety, content moderation, or policy operations at a senior level. Direct experience in LLM red-teaming/adversarial testing, documented edge-case mitigation, and localization/translation is highly preferred. Strong analytical writing and emotional resilience are essential due to the challenging nature of the content. Contributors will review, annotate, and evaluate AI-generated responses, focusing on explicit safety issues such as bias, harassment, and misinformation. Responsibilities include fact-checking, reviewing policy alignment, documenting rationales, and identifying edge cases in both Korean and English. Annotations will inform foundation-model labs, preventing unsafe or adversarial AI outputs through refined safety policy application.