About Me

Welcome! I am a PhD candidate at Université Paris Cité working on improving the efficiency of Large Language Models (LLMs) during inference. My research centres on ANN search for KV‑cache compression and LLM inference optimization. I am fortunate to be advised by Prof. Themis Palpanas.

Prior to starting my PhD, I studied Mathematics at Harbin Institute of Technology, Shenzhen and worked on pattern mining. I have a strong background in mathematics and programming (Java, C++, Python) and enjoy bridging theoretical foundations with real‑world applications.

During my MSc and subsequent research assistantships I was fortunate to collaborate with outstanding scholars. At HKUST (Guangzhou) I worked with Prof. Yuyu Luo on data selection for LLM instruction tuning. I also collaborated with Prof. Lei Chen and Dr. QiYu Liu on learned index structures. At the City University of Hong Kong, under the mentorship of Prof. Zhichao Lu, I explored multi‑objective optimisation and evolutionary computation, developing LLM‑driven automatic algorithm design (AAD). These experiences reinforced my commitment to research with real societal impact.

Education

Experience

Research Interests

• LLM inference optimization: developing efficient transformer decoding with compressed KV caches, including vector search strategies and quantisation techniques.
• Approximate nearest‑neighbour search for KV caches: designing ANN indexes tailored for the key–value memory structures used in generative models.

Selected Publications

For a complete list see my Google Scholar profile. Here are a few highlights:

• Xiaotian Lin, Yanlin Qi, Yizhang Zhu, Themis Palpanas, Chengliang Chai, Nan Tang, Yuyu Luo. LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning. arXiv preprint 2025. [link]
• Qiyu Liu^∗†, Siyuan Han^∗, Yanlin Qi^∗(co-first), Jingshu Peng, Jin Li, Longlong Lin, Lei Chen. Why Are Learned Indexes So Effective but Sometimes Ineffective? VLDB 2025. [link]
• Qiyu Liu^∗†, Yanlin Qi^∗(co-first), Siyuan Han^∗, Jingshu Peng, Jin Li, Lei Chen. Not Small Enough? SegPQ: A Learned Approach to Compress Product Quantization Codebooks. VLDB 2025.
• Sichun Luo, Bowei He, Haohan Zhao, Wei Shao, Yanlin Qi, Linqi Song. RecRanker: Instruction Tuning Large Language Model as Ranker for Top‑k Recommendation. ACM Transactions on Information Systems, 2023. [link]
• Yanlin Qi, Guoting Chen, Wensheng Gan. Mining periodic trends via closed high‑utility patterns. Expert Systems with Applications, 2023. [link]
• Yanlin Qi, Guoting Chen. Mining Valuable Fuzzy Patterns via the RFM Model. ICDMW 2022. [link]
• Xiaojie Zhang, Yanlin Qi, Guoting Chen, Wensheng Gan. Fuzzy‑driven periodic frequent pattern mining. Information Sciences, 2022. [link]
• Yanlin Qi, Guoting Chen, Wensheng Gan. F‑RFM‑Miner: An efficient algorithm for mining valuable fuzzy patterns via the RFM model. Applied Intelligence, 2023. [link]

Projects

Honours & Awards

• National Encouragement Scholarship (×3), 2018–2021.
• First‑class Scholarship, Harbin Institute of Technology (×3), 2021–2022.
• Excellent Master’s Thesis (top 2%) at Harbin Institute of Technology, Jan 2024.
• Innovation & Entrepreneurship Training Programme Award, 2019–2020.
• Excellent Student Award, Dalian Maritime University, May 2021.
• Multiple prizes in mathematics competitions at Dalian Maritime University.