模型数据运营-TikTok安全产品
Core
Design and maintain the Golden Set for AIGT (AI-generated content) safety evaluation, engineer prompts for SOTA models, and translate community guidelines into quantifiable AI Indicator Frameworks for LLM moderation.
Role type
Senior AI Safety Product Operations Engineer (LLM & Prompt Engineering)
Builds
AIGT evaluation benchmarks, risk-scenario prompt libraries, and quantifiable content safety indicator frameworks.
Domain
Content Safety / AI Safety / Large Language Models
Deliverable
production ML models
Required skills
Prompt Engineering, Few-Shot learning, RAG, SQL, data analysis, policy deconstruction, logical abstraction, cross-functional collaboration.
Preferred skills
Experience with Golden Set construction, A/B testing, model fine-tuning, English proficiency.
Technologies
LLM, SOTA models, SQL, Excel, BI tools.
Responsibilities
Plan and build the AIGT Golden Set covering core policies and content genres; design and optimize prompts for risk scenarios using SOTA models; decompose community guidelines into quantifiable AI Indicator Frameworks; identify and resolve data quality issues (dirty data, label conflicts) to support model iteration.
Seniority
Senior, hands-on IC
