Wenbin Feng

Wenbin Feng

Computer Science and Technology

About me

I am a senior undergraduate at the Big Data and Internet College, Shenzhen Technology University, majoring in Computer Science and Technology. I am advised by Prof. Yu Lu. I have received a pre-admission offer for the Master’s program at Hong Kong Polytechnic University.

My research spans computer vision and natural language processing, with a focus on multi-modal learning for medical and industrial applications.

Currently, I am an AI Agent Development Intern at MindCruise (NOTTA), working on document automation including custom template extraction and automatic generation for Word/PPT documents.

Research interests:

  • Computer Vision
  • Large Language Models
  • Reinforcement Learning
  • AI Agents (Tool-calling, ReAct, Chain-of-Thought)

Publications

  • Yu Lu*, Huilin Ge, Jie Zhu, Wenbin Feng, Jiali Ouyang, Ze Wang, Xiaoping Chen. “AquaSlot-SAM: Coupling slot-based State Space Models with SAM for robust underwater video multi-object segmentation.” Pattern Recognition (Accepted, 2025).
  • Wenbin Feng, Yu Lu*, Shijie Shi, Meng Li, Huilin Ge. “A Deep Learning Model for Surface Defect Detection in Thermoelectric Cooler Components.” Proceedings of the 21st International Conference on Intelligent Computing (ICIC 2025), Ningbo, China, July 26–29, 2025. Springer. doi:10.1007/978-981-96-9921-6_14 📄 Paper
  • Wenbin Feng, Yu Lu*, Xiaoqing Li, Shijie Shi, Yingjian Qi. “Multi-Stage Contrastive Training for Medical Report Generation: A Mamba-Based Multi-Modal Large Model.” Proceedings of the 2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Vienna, Austria, October 5-8, 2025. 📄 Paper

Internship Experience

AI Agent Development Intern | MindCruise (Notta) November 2025 - Present

  • Developing intelligent document automation systems for Notta Brain, focusing on agentic workflows for template extraction and content generation
  • Building AI-powered modules for automatic Word/PowerPoint document processing using large language models
  • Implementing tool-calling and ReAct frameworks for enhanced document understanding and generation capabilities