Wang Haojie
Assistant Researcher at the Department of Computer Science, Tsinghua University
Wang Haojie is an Assistant Researcher at the Department of Computer Science, Tsinghua University, and the Intelligent Computing Systems Research Center at Qiyuan Laboratory. His research interests include artificial intelligence systems and high-performance computing. His work has been published in top international conferences and journals such as OSDI, ATC, PPoPP, SC, PLDI, TC, and TPDS. He received the Best Student Paper Award at ICS 2021 and the Second Place Best Paper Award at TPDS 2022. He has been recognized with several honors, including the Tsinghua University Outstanding Doctoral Dissertation Award, Beijing Outstanding Graduate Award, “Shuimu Scholar” of Tsinghua University, Tsinghua University Outstanding Postdoctoral Fellow, and the ACM SIGHPC China Excellent Doctoral Dissertation Award. He has been supported by the National Natural Science Foundation of China (Youth Program) and selected for the Beijing Young Talents Program.
Topic
Unified Intelligent Computing Architecture and Open-Source Ecosystem for Heterogeneous Computing
In the context of increasing heterogeneity in intelligent computing chips and the growing difficulty of unifying the computing ecosystem, the intelligent computing software stack faces severe challenges and core demands. To address these challenges, we propose the Jiuyuan Unified Intelligent Computing Architecture, which abstracts away underlying hardware heterogeneity and provides a consistent programming model to support the construction and optimization of training and inference frameworks, effectively solving the challenges brought by heterogeneous systems. At the same time, we are actively promoting the development of an open-source ecosystem through initiatives such as open-source course development and competition-based practice, building a vibrant and collaborative intelligent computing community. Outline: 1. Research Background 2. Technical Roadmap Analysis 3. Unified Intelligent Computing Architecture for Heterogeneous Computing 4. Training and Inference Optimization Based on the Unified Intelligent Computing Architecture 5. Open-Source Community and Ecosystem Development
