shanwnwei

Spearheaded the development of the 'YunLing NMT System,' enhancing cross-border resource languages, handling over 5 million characters monthly. Pioneered the creation of a comprehensive Chinese–Vietnamese multimodal machine translation dataset, comprising approximately 50,000 parallel sentence pairs and 30,000 image-text pairs. Innovated a multimodal translation model integrating mCLIP visual embeddings with an increase in human adequacy scores on internal benchmarks. Engineered a cutting-edge lightweight defect-detection system using YOLOv5 for power-line inspections from UAV footage, achieving a remarkable [email protected] of 95% and implementing a Dockerised web dashboard for real-time analysis. Led applied research initiatives on large language models, successfully conducting pilot fine-tuning experiments to advance the field of artificial intelligence.