- Adopting Graph Attention Networks for Key Information Extraction in documents- Building interactive web-based demonstrations- Investigating and experimenting with joint detection and tracking within an end-to-end paradigm for fisheye camera surveillance

- Improved RetinaNet with Recurrent-CNN to enhance spatial features for tiny object detection in the wild.- Investigated image-based and graph-based for Key Information Extraction.- Utilized Character Region Awareness for Text Detection in Japanese documents, combined with weakly-supervised learning.- Developed models for editing text in the wild with W-GAN for data enrichment.

- Leveraged and modified deep learning models (MRCNN, EfficientNet) to develop AI solutions for defect inspection in the semiconductor industry.- Be a presenter for internal AI-related field-sharing workshops at the company.

- Investigated and implemented Convolutional Recurrent Neural Networks (CRNNs) for Vietnamese Optical Character Recognition (OCR)- Utilized state-of-the-art algorithms such as YOLO, SSD for Dense Person Detection given limited annotation data.