Brief Introduction
Hi, welcome to my homepage! I am Yan Li/李炎. I received my PhD degree in Computer Software and Theory from Software Engineering Institue, School of Computer Science, Peking University. Currently, I am a Research Engineer at Huawei. My research mainly focuses on large-scale deep learning systems, performance optimization of LLM inference engine, system-model co-design, etc.
Education
- Scholarship of Schlumberger, Dec. 2019.
- Scholarship of Schlumberger, Dec. 2020.
- Award of Scientific Research, Dec. 2017.
- Meritorious Winner of Mathematical Contest in Modeling, Apr. 2017.
- Scholarship of Freshman, Sep. 2014.
- Courses: Lab. on Operating Systems(94)/Computer Network(94)/Advanced Algebra(92)/Lab. on Compiler Design(91)/Mathematical Logic(91)/Mathematical Analysis(90)/etc.
Experiences
Projects
Here are some of the projects I participated in before.
SamProf - An application-oblivious profiler for cloud-hosted deep learning workloads.
Docklet - A cloud operating system for VPC (virtual private cloud).
Akame - A high-available and cost-efficient cloud-of-clouds storage service.
Publications
Here are some of my publications.
Speculative MoE: Communication Efficient Parallel MoE Inference with Speculative Token and Expert Pre-scheduling
arxiv
Centrum: Escape from the Gaussian Process World! Enhancing Database Auto-tuning with Tree-Ensemble Bayesian Optimization
SIGMOD 2025
Performance Modeling for Cloud-hosted Deep Learning Services with Hybrid Representation
arxiv
SamProf: Top-down Performance Analysis for Neural Networks via Instruction Sampling
axriv
Sectum: Accurate Latency Prediction for TEE-hosted Deep Learning Inference
The 42th IEEE International Conference on Distributed Computing Systems (ICDCS 2022), Bologna, Italy, Jul 10-Jul 13, 2022
SpotTune: Leveraging Transient Resources for Cost-efficient Hyper-parameter Tuning in the Public Cloud
The 40th IEEE International Conference on Distributed Computing Systems (ICDCS 2020), Singapore, Nov 29-Dec 1, 2020
PrTaurus: An Availability-Enhanced EMR Service on Preemptible Cloud Instances
2020 IEEE International Conference on Web Services (ICWS 2020). IEEE, Beijing, China, Oct 18-24, 2020
Cross-domain Workloads Performance Prediction via Runtime Metrics Transferring
The 11th International Workshop on Joint Cloud Computing (JCC2020), Oxford, UK, Aug 3-6, 2020
CloudMeter: A Tool To Select the Best Cloud Service
The 11th International Workshop on Joint Cloud Computing (JCC2020), Oxford, UK, Aug 3-6, 2020
DCStore: A Deduplication-Based Cloud-of-Clouds Storage Service
IEEE International Conference on Web Service (ICWS 2019), Milan, Italy, Jul 8-13, 2019
Comparison between Chunk-based and Layer-based Container Image Storage Approaches: an Empirical Study
The 10th International Workshop on Joint Cloud Computing (JCC2019), San Francisco, USA, Apr 4-9, 2019
GPU Scheduling for Short Tasks in Private Cloud
The 10th International Workshop on Joint Cloud Computing (JCC2019), San Francisco, USA, Apr 4-9, 2019
BDViewer - A Web-Based Big Data Processing and Visualization Tool
IEEE International Conference on Computers Software and Applications (COMPSAC2018), Tokyo, Japan, Jul 23-27, 2018