Jie Hu (胡婕)

Hey, I'm Jie Hu, a PhD candidate at the Center for Data Science, Peking University, where I am incredibly fortunate to be advised by Prof. Zaiwen Wen and Prof. Kun Yuan. Previously, I obtained my B.S. in Applied Mathematics from Tongji University in 2023.

My research lies at the intersection of machine learning algorithms and systems, focusing on efficient and scalable large language models. I'm broadly interested in efficient inference, agentic systems, RL infrastructure, and model–system co-design. I'm currently a research intern at Tencent Hunyuan, working on algorithms and systems for efficient large-scale LLM deployment.

Feel free to reach out via email—always happy to connect and exchange ideas!

• 2026.09 Started as a research intern at Tencent Hunyuan, working on algorithms and systems for efficient large-scale LLM deployment.
• 2026.09 TierKV is accepted to EuroSys 2027 — long-context on-device LLMs via predictive multi-tier KV caching.
• 2026.05 DFSAttn is accepted to ICML 2026 — dynamic fine-grained sparse attention for faster video generation.
• 2026.01 Passed the doctoral qualifying exam and became a PhD candidate.
• 2025.07 Wrapped up a one-year research internship at Huawei 2012 Lab.
• 2025.06 CentroidKV is on arXiv, and later accepted to TMLR — online KV cache clustering for long-context LLM inference.
• 2024.05 Our work Distributed Bilevel Optimization with Communication Compression is accepted to ICML 2024.
• 2023.09 Started my PhD at the Center for Data Science, Peking University.

For the full and most up-to-date list, see my Google Scholar profile.

• 2026 Fall at PKU Foundation Models: Theory and Applications
  Teaching Assistant
• 2025 Spring at PKU Introduction to Foundation Models
  Teaching Assistant