Zhi Wang (王智)
Associate Professor
Room 1708, Information Building
Phone: +86-755-26031113 |
|
Research
With the rapid rise of large-scale pretrained models and autonomous intelligence, AI is increasingly moving from the digital realm into the physical world, catalyzing a new industrial transformation centered on industrial foundation models and embodied industrial intelligence. At mmlab@SIGS, we focus on cutting-edge interdisciplinary research at the intersection of autonomous multimedia intelligence, distributed machine learning, and industrial foundation models with embodied intelligence. Targeting real-world industrial settings, we investigate end-to-end intelligent systems spanning data perception, knowledge modeling, and decision-to-action execution. Building on our long-standing theoretical and technical foundations in large-scale networked multimedia systems, we extend the established framework of “cross-domain perception, integrated scheduling, and cloud–edge collaboration” into a new industrial paradigm characterized by “autonomous perception, intelligent generation, and self-organizing topology,” with the goal of enabling tightly integrated “perception–cognition–execution” industrial intelligence.
The group has achieved internationally recognized results in related areas, including multiple national and provincial-level science and technology awards, high-impact publications in leading venues such as NeurIPS, SIGCOMM, MobiCom, CVPR/ICCV, and ACM Multimedia, and demonstrated academic influence. Several core technologies have been deployed in mission-critical applications at major enterprises and further advanced through technology transfer and startup incubation. Supported by Tsinghua SIGS research platforms and strong computational resources, we provide students with a complete research pipeline covering large-model training, simulation-based validation, and practical industrial deployment. Through deep industry–academia collaboration and a dual-mentorship practice system, we aim to cultivate future leaders with both rigorous theoretical depth and strong engineering capabilities, and we welcome motivated students eager to pursue frontier research with real-world impact in industrial foundation models, embodied intelligence, and distributed learning.
- Autonomous multimedia intelligence
- Distributed machine learning
- Industrial foundation models
- Embodied industrial intelligence
Research Groups
- Efficient Deep Learning & Embodiment→ Model compression, efficient neural architectures, and deploying deep learning on resource-constrained embodied systems for real-time robotic perception and control.
- Agentic & Embodied Large Model Training→ Training and accelerating large vision-language-action models for agentic decision-making, with focus on scalable distributed training and embodied AI deployment.
- Self-Evolving Embodied Agents→ Autonomous agents that continuously self-improve through interaction with physical environments, leveraging world models and reinforcement learning for open-world adaptation.
Join Us
Selected Honors and Awards full list →
- Best Paper Award Finalist, ACM Multimedia 2025
- 北京市科学技术奖自然科学奖二等奖, 2024
- Grand Challenge 1st Place, V3Det Challenge, CVPR, 2024
- Best Paper Award, IEEE Transactions on Multimedia, 2022
- Best Paper Award, ACM Multimedia, HUMA Workshop 2021
- Best Paper Award Finalist, IEEE ICME 2021
- 国家级青年人才, 2020
- 国家自然科学二等奖, 2018
Selected Publications (2023–) full list →
Teaching & Education
Courses
- 高级大数据系统（MOOC） — 首批国家级一流本科课程、国家精品在线开放课程 | 工程类专业学位研究生在线示范课程
- 大数据系统基础B（64100033），秋季；2014–
- 大数据科学与应用系列讲座（60250131），秋季；2015– — 深圳市"精品共享课"（2024）
- 分布式机器学习（85990072），春季；2021– — 清华大学教评前5%（2022学年）
- 文献检索与论文写作（82558001），春季；2021–
- 大数据可视化（80240683），春季/秋季；2016–2019
- 大数据分析基础（86000152），春季/秋季；2015–2017
Textbook
- 参编，《PLC应用技术（西门子s7-1200）》，西北工业大学出版社（CIP 数据核字第2025CD3335号）
Teaching Awards
- 优秀指导教师，首届全国大学生职业规划大赛北京市赛就业赛道，2024
- 优秀指导教师，首届清华大学大学生职业规划大赛就业赛道一等奖，2023
- 清华大学深圳国际研究生院年度优秀，2023
- 教育部在线教育中心智慧教学之星，2019
- 清华大学先进工作者，2017
- 教育部在线教育中心教学先锋奖励，2016
Selected Patents full list →
- 模型优化方法、电子设备以及计算机可读存储介质； 2026（SIGS A级核心高价值专利）
- 基于强化学习的动态多媒体数据部署方法； 2025（SIGS A级核心高价值专利）
- 一种用于生成图像的扩散模型混合精度量化方法； 2025
- 一种用于边缘设备的无需重训练量化领域自适应方法； 2025
- 基于强化学习的联合决策方法及装置； 2023
- 一种基于强化学习的图片动态自适应压缩方法； 2021
- 基于数据分布的机器深度学习方法； 2020
- 一种基于迁移学习的车流量预测方法和系统； 2022