多模态算法工程师-抖音评论
Core
Develop large model-driven automated review systems for safety, experience, growth, and innovation of interactive content (comments, danmu) across video, live streaming, and image-text formats on Douyin.
Role type
Senior multimodal algorithm engineer (content safety & understanding)
Builds
Large model-driven automated review systems, unified content understanding technology, and innovative interactive content products (e.g., smart summaries, host assistants).
Domain
Social media, content moderation, multimodal AI
Deliverable
production ML models
Required skills
CV, VLM, MLLM, deep learning training frameworks, mathematical foundations, LLM/MLLM pre-training and post-training (SFT, RL), Auto-prompt Engineering, Embedding, In-context learning, RAG
Preferred skills
Experience in short video/image/live streaming algorithms, publications in top-tier conferences (NIPS, ICML, CVPR, etc.), competitive programming experience
Technologies
LLM, MLLM, SFT, RL, Auto-prompt Engineering, Embedding, In-context learning, RAG
Responsibilities
Research and develop large model-driven systems to address semantic and structural challenges in interactive content; Analyze user intent to optimize ranking and distribution strategies; Explore innovative product forms to improve interaction efficiency and user growth; Optimize general large models for content safety and understanding.
Seniority
Senior, hands-on IC
