I've developed an AI-based image-text retrieval system using CLIP for cross-modal search and semantic matching, using ViT-B/32 and the Flickr8k dataset.
For my master's project, I applied Grad-CAM and patch-level similarity analysis to make retrieval results more interpretable, and evaluated performance with Recall@K metrics. I've also supported Nanjing University of Information Science and Technology's student portal through bug fixes, functional improvements, database query optimization, and React web applications.
At Scale Force in France, I built responsive, interactive web pages and helped improve website usability and functionality across web-based projects.
My work also includes complete branding directions for Vexaled and Decora, spanning visual identities, brand guidelines, product communication, advertising concepts, social media content, and B2B marketing materials.
