多模态大模型算法（国际化直播） - TikTok算法
Core
Developing and deploying multimodal large language models (LLMs) to optimize live streaming recommendations, content representation, and safety governance on TikTok.
Role type
Senior IC multimodal LLM algorithm engineer (live streaming)
Builds
Next-generation LLM recommendation architectures (LLM4Rec, Agentic Rec), semantic ID systems, and unified base models for recommendation and governance.
Domain
Social media, live streaming, large-scale recommendation systems, generative AI
Deliverable
production ML models
Required skills
Multimodal understanding, large language model architecture and training, reinforcement learning, agent systems, experimental analysis, end-to-end model deployment
Preferred skills
Publications in top-tier conferences (NeurIPS, CVPR, ICCV, ICML), end-to-end content understanding in recommendation/search/governance
Technologies
LLMs, Multi-Agent systems, Agentic RL, CPT (Continual Pre-training), semantic ID
Responsibilities
Architect and scale LLM4Rec and Agentic Rec for live streaming; build semantic ID systems for real-time content representation; apply post-training and RL to safety governance; lead base model construction including data pipelines and pre-training; define research directions aligned with live streaming scenarios.