大模型内容安全产品 - AI创新业务
Core
Design and optimize content safety strategies for the Doubao large language model across user inputs, model reasoning, and generated outputs to balance safety with user experience.
Role type
Senior Product Manager, Large Model Content Safety
Builds
Content safety strategy systems for LLMs covering text, images, multi-turn dialogue, VLM, and real-time audio/video
Domain
AI / Large Language Models / Content Safety
Deliverable
production ML models
Required skills
LLM safety strategy design, risk analysis (false positives/negatives), SQL, prompt engineering, RLHF/SET concepts, multi-modal safety (VLM, audio/video), stakeholder management
Preferred skills
Experience with UGC moderation, political/pornographic risk frameworks, incident response, cross-functional team leadership
Technologies
SQL, LLMs, VLM, RLHF, SET
Responsibilities
Monitor and analyze safety risks and negative cases to optimize rules and models; Design safety strategies for complex scenarios like multi-turn dialogue and real-time calls; Collaborate with training teams to convert risks into training data and evaluation sets; Evaluate and approve safety integration for new features and businesses.
