• Junhao Hu
  • Publications
  • Awards
  • Services
  • Work Experience
  • Secrets
    Junhao Hu

    Junhao Hu

    Third-year Ph.D. student at CS@PKU, interested in applications and system optimizations of Large Language Models (LLMs).

    • Beijing, China
    • Email
    • Twitter
    • Github
    • Google Scholar

    • Sep. 2025 - Present: Research Assistant in Xiaomi top talent (on par with ByteDance TopSeed, Tencent Qingyun, etc.), with Fuli Luo.
      • Lead the Model architecture and infra co-design project.
    • July 2024 - Aug. 2025: Research Assistant in Huawei Cloud, with Yizhou Shan.
      • Lead the Positional-Independent Caching (PIC) Project, which leads to a paper accepted by ICML 2025.
      • Lead the Reasoning-Aware Attention Sparsity (RaaS) Project, which leads to a paper accepted by ACL 2025.
      • Co-Lead the DeepFlow Project, which leads to a paper accepted by ATC 2025.
    • Mar. 2022 - June 2024: Research Assistant in WXG of Tencent Inc, with Yuetang Deng and Hailiang Huang
      • Lead the adaptive distributed build Project, which leads to a paper accepted by ASE 2023.
      • Co-lead the AI-assisted code generation Project, which leads to a paper accepted by ESEC/FSE 2023.
    • Dec. 2021 - Feb. 2022: Engineer intern in Data/AML of ByteDance, with Xuan Zou.
      • Rewrite TensorFlow models using ByteDance’s internal framework.
    Sitemap
    • Follow:
    • GitHub
    • Feed
    © 2025 Junhao Hu. Powered by Jekyll & AcademicPages, a fork of Minimal Mistakes.