Content Understanding Multimodal Model Algorithm Engineer-Global E-commerce-Soaring Star Talent Program

Build and research multimodal large language models for intelligent e-commerce content moderation. Work on modality fusion for text/audio/image/video/live data, strengthen multilingual long-sequence few-shot and zero-shot capabilities, and develop adversarial defense for AIGC-generated content. Enable RAG-based case retrieval, tool invocation, and agent planning to decompose PBR changes into atomic moderation tasks, aiming for interpretable COT and high-accuracy autonomous review.