Zeyu Mi
Associate Professor
Works on large model systems and operating systems, including the PowerInfer on-device LLM inference engine.
Zeyu Mi's lab at SJTU works on efficient training & inference, distributed, cloud & edge systems and large language models. 14 papers since 2023, 4 at top venues, mostly OSDI and SOSP. 4 PhD students and 7 master's students. 14 alumni. Applications close .
At a glance
- papers, 2023–2026
- 4/ year
- at T1 venues
- 29%
- PhD students
- 4
- MS students
- 7
- alumni listed
- 14
Papers by yearfrom the lab homepage, 2023–2026
- 6 papers2026 · T1 0 · collecting
- 1 papers2025 · T1 0
- 4 papers2024 · T1 2
- 3 papers2023 · T1 2
T1 venuesOtherVenues: OSDI, SOSP
Research focuspaper titles that use each topic's words
- Efficient training & inference5 papers
- Large language models5 papers
- Distributed, cloud & edge systems0 papers
- large model0 papers
- operating systems0 papers
Paper list
From the lab homepage.
20266 papers
- The ACM International Conference on Architectural Support for Programming Languages and Operating Systems, March .DIP: Addressing Data Dynamicity in Large Multimodal Model Training with Dynamic Interleaved Pipeline
- IEEE Transactions on Dependable and Secure ComputingEnabling Confidential VMs Across Diverse ARM Platforms by Retrofitting TrustZone
- Proceedings of the ACM SIGOPS Annual Technical Conference (ATC'26), Shatin, Hong Kong, China, NovemberHeadFlow: Scaling Long-Context Training with Head-Group Activation Residency
- In Proceedings of the USENIX Symposium on Operating Systems Design and Implementation (OSDI '26), JulyInference in the Shadows: Taming Memory Bandwidth Contention in Mobile LLM Inference with Sereno
- IEEE Transactions on ComputersNeuronInfer: Fast Large Language Model Inference on a Smartphone
- The USENIX Conference on File and Storage Technologies, February .SolidAttention: Low-Latency SSD-based Serving on Memory-Constrained PCs
20251 paper
- The ACM International Conference on Architectural Support for Programming Languages and Operating Systems, March .PipeLLM: Fast and Confidential Large Language Model Services with Speculative Pipelined Encryption
20244 papers
- USENIX Annual Technical Conference, JulyCPC: Flexible, Secure and Efficient CVM Maintenance with Confidential Procedure Calls
- The ACM International Conference on Architectural Support for Programming Languages and Operating Systems, AprilCPS: A Cooperative Para-virtualized Scheduling Framework for Manycore Machines
- SOSP (T1)PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPU
- SOSP (T1)VPRI: Optimized I/O Page Fault in Public IaaS
20233 papers
- USENIX Annual Technical Conference, JulyAnalysis and Optimization of Network I/O Tax in Confidential Virtual Machines
- OSDI (T1)Encrypted Databases Made Secure Yet Maintainable
- OSDI (T1)Security and Performance in the Delegated User-level Virtualization
Where alumni went14 listed
Apply to SJTUFall 2027
Computer Science PhD (international applicants)
- Round
- Self-funded
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUW
Baris Kasikci
- Efficient training & inference
- USUCSD
Yiying Zhang
- Efficient training & inference
- SGNUS
Bingsheng He
- Accelerators, chips & architecture
- CNZhejiang
Xiaohang Wang
- Accelerators, chips & architecture
- CNPeking
Xin Jin
- Efficient training & inference
- USNYU
Hongyi Wen
- Recommendation, search & web
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
SJTUcloses