Boyang He

• Proficient in Python programming, familiar with Java, and knowledgeable about common data structures; possesses strong engineering skills and awareness of coding standards.
• Skilled in mainstream deep learning frameworks such as PyTorch and TensorFlow, with practical experience in model building, training, tuning, and deployment; basic experience with Docker; and capable of configuring environments and troubleshooting issues under Linux.
• Solid understanding of computer vision fundamentals, including CNNs, YOLO object detection, image enhancement, and image classification; possesses comprehensive hands-on project experience.
• Familiar with the Transformer, encoder-decoder architecture, and self-attention mechanism; understands fundamental theories of pre-training models such as BERT and GPT; has a solid theoretical foundation for further NLP applications.
• Strong English reading and writing skills (IELTS 7.0, GRE 321), able to efficiently read technical documents, and communicate in English for daily work.
• Strong sense of responsibility and execution; able to efficiently complete tasks in a team environment; good team collaboration skills.
• Strong logical thinking, learning ability, and curiosity for new technologies.

Led deep learning model implementation (ResNet18) for land cover classification in remote sensing images; integrated interpretability methods (Grad-CAM), achieving F1-score of 0.97.

Developed a deep learning system for automatic detection of prohibited items (e.g., knives) using YOLOv8 and EfficientDet; achieved mAP 0.89 on occluded test sets, surpassing baseline methods.

Developed multiple independent projects using Python, Java, and web technologies; built GUIs, interactive systems, and participated in 3D space modeling with Unreal Engine.