Han Zhao
Assistant Professor
Works on DNN and LLM inference and training systems, datacenter resource management and serverless architectures.
Han Zhao's lab at SJTU works on efficient training & inference and accelerators, chips & architecture. 21 papers since 2023, 9 at top venues, mostly ASPLOS and EuroSys. 2 PhD students and 5 master's students. 2 alumni. Applications close .
At a glance
- papers, 2023–2026
- 5/ year
- at T1 venues
- 43%
- PhD students
- 2
- MS students
- 5
- alumni listed
- 2
Papers by yearfrom the lab homepage, 2023–2026
- 8 papers2026 · T1 7 · collecting
- 6 papers2025 · T1 1
- 5 papers2024 · T1 1
- 2 papers2023 · T1 0
T1 venuesOtherVenues: ASPLOS, EuroSys, HPCA, MICRO, NeurIPS
Research focuspaper titles that use each topic's words
- Efficient training & inference7 papers
- Accelerators, chips & architecture3 papers
- training systems0 papers
- datacenter0 papers
- dnn0 papers
Paper list
From the lab homepage.
20268 papers
- EuroSys (T1)Arena: Efficiently Training Large Models via Dynamic Scheduling and Adaptive Parallelism Co-Design
- MICRO (T1)Deltoris: Enabling Real-time VLA Inference in Embodied Al via Bit-level Sparsity and Speculative Inference
- NSDI (T1)Flare: Anomaly Diagnostics for Divergent LLM Training in GPU Clusters of Thousand-Plus Scale
- TACOFrom Overloaded to Guaranteed: High-Throughput Multi-SLO Enforcement for LoRA-Assisted On-Premise LLM Deployment
- NeurIPS (T1)Justitia: Fair and Efficient Scheduling of Task-parallel LLM Agents with Selective Pampering
- HPCA (T1)LEGO: Supporting LLM-enhanced Games with One Gaming GPU
- ASPLOS (T1)Towards High-Goodput LLM Serving with Prefill-decode Multiplexing
- HPCA (T1)Towards Resource-Efficient Serverless LLM Inference with SLINFER
20256 papers
- TACOARACHNE: Optimizing distributed parallel applications with reduced inter-process communication
- APPTDACO: Unlocking Latent Dataflow Opportunities in Edge-Side SIMT Accelerators
- TACOEDAS: Enabling Fast Data Loading for GPU Serverless Computing
- ATCEfficient Performance-Aware GPU Sharing with Compatibility and Isolation through Kernel Space Interception
- EuroSys (T1)Improving GPU Sharing Performance through Adaptive Bubbleless Spatial-Temporal Sharing
- TACOTaming Flexible Job Packing in Deep Learning Training Clusters
20245 papers
- TCAdaptive Kernel Fusion for Improving the GPU Utilization while Ensuring QoS
- Chinese Science Information ScienceExploiting all intra-SM parallelism to maximize the throughput while ensuring QoS
- ASPLOS (T1)FaaSMem: Improving Memory Efficiency of Serverless Computing with Memory Pool Architecture
- FCSFLAPS: fluctuation-aware power auction strategy for reducing the power overload probability
- TACOPotamoi: Accelerating neural rendering via a unified streaming architecture
20232 papers
- TCImproving Cluster Utilization Through Adaptive Resource Management for Deep Neural Network and CPU Jobs Colocation
- SoCCMaximizing the Utilization of GPUs Used by Cloud Gaming through Adaptive Co-location with Combo
Where alumni went2 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
- CNPeking
Xuanzhe Liu
- Distributed, cloud & edge systems
- CNSJTU
Zihan Wu
- Accelerators, chips & architecture
- SGNTU
Chen Wang
- Distributed, cloud & edge systems
- CHETH Zurich
Ana Klimovic
- Efficient training & inference
- USUSC
Ibrahim Sabek
- Data management & big data
- CNSJTU
Qinya Li
- Distributed, cloud & edge systems
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