Wenjun Wang is a researcher at the Hong Kong Polytechnic University specialising in efficient multimodal large language models. Wenjun holds a doctorate and focuses on developing innovative techniques to reduce computational costs while maintaining high performance. His research includes dynamic sparse attention mechanisms, adapter-based modular fusion, and hardware-aware strategies such as hybrid-precision quantisation and curriculum contrastive learning. Wenjun’s work has achieved notable results, including significant memory reduction and faster inference with minimal accuracy loss. He is currently exploring energy-efficient neuromorphic architectures and real-time embodied AI interfaces. By integrating algorithmic advances with system-level optimisations, Wenjun aims to bridge theory and practical deployment, advancing scalable and efficient AI applications.