📋 Job Overview

The role involves developing multimodal large models for security applications in WeChat, focusing on modalities such as images, videos, behavioral sequences, and relational features to understand content and industry rules for auditing and risk control. Responsibilities include optimizing algorithms for practical deployment through data processing acceleration, training efficiency improvements, and inference speed enhancements. The position also entails exploring cutting-edge multimodal technologies like reinforcement learning, domain knowledge injection, cross-modal alignment, and pre-training for security applications.