👋 I am an M.S. candidate in Artificial Intelligence at Xidian University, Xi'an, China (graduating July 2026). My research focuses on open-vocabulary object detection and multi-object tracking robustness under semantic distribution shift — the gap between what vision-language models are trained to recognize and what they encounter at deployment.

I am the first author of EnYOLO-World, which improves YOLO-World with PGI and FiLM-PAN (+3.1 AP, 52.2 AP on COCO, 78 FPS, under review at ELCVIA), and lead researcher on VocabDriftMOT, the first benchmark for vocabulary-drift failure modes in open-vocabulary multi-object tracking. Beyond research, I have 4 years of production software engineering experience and have shipped systems using PyTorch, Docker, GCP Cloud Run, FastAPI, and ChromaDB.

Experience

Featured Projects