2025
Zhipu AI, China
GLM-4 Pretraining Algorithm Internship
Zhipu AI, Dec. 2024 - Present
Focused on pretraining of GLM-4 models, leveraging techniques such as Chain-of-Thought (CoT) to enhance the model's ability to perform mathematical reasoning tasks. Research and reproduce the qwen and DeepSeek algorithms. Worked on improving the model's performance in solving complex mathematical problems by refining training strategies and optimizing the reasoning process.
Zhipu AI, China
Yu Zhang, Aohan Zeng, Jie Tang
Zhipu AI 2025
Focused on pretraining of GLM-4 models, leveraging techniques such as Chain-of-Thought (CoT) to enhance the model's ability to perform mathematical reasoning tasks. Research and reproduce the qwen and DeepSeek algorithms. Worked on improving the model's performance in solving complex mathematical problems by refining training strategies and optimizing the reasoning process.
2024
Shanghai AI Laboratory Large Chip Group
Research Assistant, Algorithm Technical Leader
Shanghai AI Laboratory, Jun. 2023 - present
Working under Professor Shouyi Yin, this project explores compression technologies for general visual models, with two papers under preparation for submission to CVPR 2025 and ICCV 2025. Developed a new adaptive knowledge distillation algorithm, achieving high-quality model compression with an accuracy reduction of less than 2%.
Shanghai AI Laboratory Large Chip Group
Yu Zhang, Shujun Peng, Xinhan Lin#, Shouyi Yin#, Yang Hu# (# corresponding author)
Shanghai AI Laboratory 2024
Working under Professor Shouyi Yin, this project explores compression technologies for general visual models, with two papers under preparation for submission to CVPR 2025 and ICCV 2025. Developed a new adaptive knowledge distillation algorithm, achieving high-quality model compression with an accuracy reduction of less than 2%.
China National Nuclear Corporation (CNNC) Energy Technology Co., Ltd., China
Technology Management Intern
CNNC, Jun. 2024 - Jan. 2025
Served as the technical leader of the **National Key R & D Project**. Deeply involved in the High-Temperature Reactor Technology Collaboration and Optimization Project at CNNC. Responsible for data collection and processing, and assisted in constructing a predictive model for the development of high-temperature reactors. Contributed to the production of a detailed roadmap report for the future development of high-temperature reactors.
China National Nuclear Corporation (CNNC) Energy Technology Co., Ltd., China
Yu Zhang, Wei Xiong, Jing Chen, Guangduo Xu
CNNC 2024
Served as the technical leader of the **National Key R & D Project**. Deeply involved in the High-Temperature Reactor Technology Collaboration and Optimization Project at CNNC. Responsible for data collection and processing, and assisted in constructing a predictive model for the development of high-temperature reactors. Contributed to the production of a detailed roadmap report for the future development of high-temperature reactors.
Mercedes-Benz Automotive Co., Ltd., China
Machine Learning and Digital Twin Algorithms Internship
Mercedes-Benz, May 2024 - Oct. 2024
Participated in the Carbon Neutrality Capacity Enhancement Program at Tsinghua University, while interning at Beijing Benz Automotive Co., Ltd. Focused on the development of a building-level integrated energy system based on digital twin technology. Contributed to energy-saving initiatives through the design and application of digital twin and machine learning algorithms.
Mercedes-Benz Automotive Co., Ltd., China
Yu Zhang, Siming Ji, Yang Zou, Xiaomin Chang
Mercedes-Benz 2024
Participated in the Carbon Neutrality Capacity Enhancement Program at Tsinghua University, while interning at Beijing Benz Automotive Co., Ltd. Focused on the development of a building-level integrated energy system based on digital twin technology. Contributed to energy-saving initiatives through the design and application of digital twin and machine learning algorithms.
Software Outsourcing Studio
Project Founder
Independent, Oct. 2020 - present
Founded a student-led software outsourcing studio, providing software solutions for businesses and educational institutions, achieving over 600,000 RMB in revenue in 2021. Co-founded Nanjing Zhiyu Information Technology with a capital of 12.5 million RMB, specializing in IoT security products.
Software Outsourcing Studio
Yu Zhang
Independent 2020
Founded a student-led software outsourcing studio, providing software solutions for businesses and educational institutions, achieving over 600,000 RMB in revenue in 2021. Co-founded Nanjing Zhiyu Information Technology with a capital of 12.5 million RMB, specializing in IoT security products.
2023
Peng Cheng National Key Laboratory, Tsinghua University
Research Assistant, Algorithm Technical Leader
Tsinghua University, Mar. 2023 - Jun. 2024
Collaborated with Professor Qing Li's research team at Tsinghua Shenzhen International Graduate School, focusing on diverse network attack detection and anomaly detection for edge devices. Developed a decision-tree-based unsupervised anomaly detection model, optimized for lightweight efficiency, versatility, and high-performance frameworks. Achievements include three publications (NIPS 2023 as second author, INFOCOM as second author, and NIPS 2024 as first author).
Peng Cheng National Key Laboratory, Tsinghua University
Yu Zhang, Ruoyu Li, Qing Li#, Yong Jiang# (# corresponding author)
Tsinghua University 2022
Collaborated with Professor Qing Li's research team at Tsinghua Shenzhen International Graduate School, focusing on diverse network attack detection and anomaly detection for edge devices. Developed a decision-tree-based unsupervised anomaly detection model, optimized for lightweight efficiency, versatility, and high-performance frameworks. Achievements include three publications (NIPS 2023 as second author, INFOCOM as second author, and NIPS 2024 as first author).
2022
Deadline-Aware Dynamic Task Scheduling in Edge–Cloud Collaborative Computing
Project Leader
Electronics, Jun. 2021 - Aug. 2022
This paper explores a dynamic task scheduling approach in edge-cloud collaborative computing to meet the requirements of modern cyber-physical systems and Industrial IoT applications. The proposed scheduling algorithm, DSOTS, optimizes server selection and reduces latency in time-sensitive tasks. Further enhancement through a greedy strategy, TSGS, achieves server load balancing and minimizes SLA violations. Simulation results in an edge-cloud environment indicate that DSOTS reduces average processing time and cost by 30% and 45%, respectively, and SLA violations by 25% over current solutions.
Deadline-Aware Dynamic Task Scheduling in Edge–Cloud Collaborative Computing
Yu Zhang, Bing Tang, Jincheng Luo, Jiaming Zhang
Electronics 2022
This paper explores a dynamic task scheduling approach in edge-cloud collaborative computing to meet the requirements of modern cyber-physical systems and Industrial IoT applications. The proposed scheduling algorithm, DSOTS, optimizes server selection and reduces latency in time-sensitive tasks. Further enhancement through a greedy strategy, TSGS, achieves server load balancing and minimizes SLA violations. Simulation results in an edge-cloud environment indicate that DSOTS reduces average processing time and cost by 30% and 45%, respectively, and SLA violations by 25% over current solutions.
Task Scheduling Optimization in Edge-Cloud Computing for Smart Factories
Project Leader
Hunan University of Science and Technology, Mar. 2021 - Aug. 2022
Led a research project focusing on optimizing task scheduling for time-sensitive tasks in cloud-edge collaborative environments. Developed a deadline-aware, dynamic scheduling algorithm, improving efficiency by 18%-25% and successfully implemented in Guangdong Mobile Company’s services.
Task Scheduling Optimization in Edge-Cloud Computing for Smart Factories
Yu Zhang, Tang Bing# (# corresponding author)
Hunan University of Science and Technology 2022
Led a research project focusing on optimizing task scheduling for time-sensitive tasks in cloud-edge collaborative environments. Developed a deadline-aware, dynamic scheduling algorithm, improving efficiency by 18%-25% and successfully implemented in Guangdong Mobile Company’s services.
2021
Office-Tab-for-Mac
Project Founder
By my self-development, Nov. 2022 - Dec. 2022
I have developed an Office-tab plug-in for Mac users to help Microsoft Office users control the Office products on their Mac computers for tabbed multi-file management. Microsoft Office users currently face inefficiencies when managing multiple Office files as each opens in a separate window. Implementing a tab-based interface within Office applications significantly improves multi-file navigation and enhances productivity.
Office-Tab-for-Mac
Yu Zhang
By my self-development 2022
I have developed an Office-tab plug-in for Mac users to help Microsoft Office users control the Office products on their Mac computers for tabbed multi-file management. Microsoft Office users currently face inefficiencies when managing multiple Office files as each opens in a separate window. Implementing a tab-based interface within Office applications significantly improves multi-file navigation and enhances productivity.
Software and Web Development for Zhuzhou Gear Company
Project Leader
Zhuzhou Gear Company, Feb. 2021 - Nov. 2021
Led the team in developing a gear status analysis model and web application, predicting operational status based on various parameters. Utilized a Vue and SpringBoot stack with MATLAB algorithms for real-time computation and expert system integration for data visualization.
Software and Web Development for Zhuzhou Gear Company
Yu Zhang, Zhigang Wang, Jiawei Nie, Jianjun Long, Qiongbing Zhang#, Lei Jiang#, Sha Wu# (# corresponding author)
Zhuzhou Gear Company 2021
Led the team in developing a gear status analysis model and web application, predicting operational status based on various parameters. Utilized a Vue and SpringBoot stack with MATLAB algorithms for real-time computation and expert system integration for data visualization.
Spatiotemporal Trajectory Data Storage and Indexing Optimization
Project Leader
Hunan Education Department,
Funded by the Hunan Education Department, this project improved data indexing efficiency in intelligent transportation systems by over 10% using KD-tree and NN-based algorithms. Led all stages from requirements analysis to implementation on the Spark platform, receiving a provincial grant of 50,000 RMB.
Spatiotemporal Trajectory Data Storage and Indexing Optimization
Yu Zhang, Bing Tang# (# corresponding author)
Hunan Education Department 2021
Funded by the Hunan Education Department, this project improved data indexing efficiency in intelligent transportation systems by over 10% using KD-tree and NN-based algorithms. Led all stages from requirements analysis to implementation on the Spark platform, receiving a provincial grant of 50,000 RMB.
2020
Exam Paper Generation System Development
Project Leader
University Project, Sept. 2020 - Aug. 2023
Developed an end-to-end exam paper generation system for educational informatization, covering all stages from paper creation to grading. Employed SpringCloudAlibaba for a distributed microservice framework, supporting web, mobile, and desktop platforms.
Exam Paper Generation System Development
Yu Zhang, Zhigang Wang, Jiawei Nie, Jianjun Long, Qiongbing Zhang#, Lei Jiang#, Sha Wu# (# corresponding author)
University Project 2020
Developed an end-to-end exam paper generation system for educational informatization, covering all stages from paper creation to grading. Employed SpringCloudAlibaba for a distributed microservice framework, supporting web, mobile, and desktop platforms.