模型数据运营（评估与研判中心）-TikTok安全产品
Core
Define golden sets and quantitative SOPs for LLM safety across risk domains (pornography, violence, politics, harassment); design rule routing logic for LLM execution; build and manage an AI content quality management system.
Role type
Senior LLM Model Operations Engineer (Safety & Evaluation)
Builds
LLM training datasets, safety evaluation pipelines, and rule routing logic for content moderation
Domain
AI Safety / Large Language Models / Content Moderation
Deliverable
production ML models
Required skills
LLM model training data annotation, model iteration effect evaluation, quantitative SOP creation, rule routing logic design, AI content quality management, data analysis, risk assessment
Preferred skills
Search/AI product or technical background, experience with ChatGPT/Gemini training, community rule formulation, stakeholder collaboration
Technologies
LLM frameworks, data annotation platforms, quality assurance tools
Responsibilities
Define golden sets and quantitative standards for risk domains; manage ground truth delivery including analysis, customization, training, and QC; build internal case adjudication mechanisms; design and optimize LLM rule routing logic; diagnose quality issues and drive model optimization.
Seniority
Senior, hands-on IC
