profile picture

Zishan Shao

Ph.D Student at Duke University
zishan.shao@duke.edu

© Zishan Shao 2025

Duke ECE · CEI

Hi, I’m Zishan Shao.

I am a Ph.D. student in Electrical and Computer Engineering at Duke University, advised by Prof. Yiran Chen. I develop efficient and interpretable techniques for multimodal generative models, low-rank methods, and AI algorithm–system co-design.

MLSys Efficient AI Multimodal AI
ICML 2026 · Spotlight

DecodeShare

Identifies a compact, task-general subspace in decode-time LLM states and tests its causal role with matched interventions.

AAAI 2026 · Poster

FlashSVD

Pairs a rank-aware CUDA/Triton kernel suite with a unified runtime for memory-efficient low-rank Transformer inference.

ACM MM 2026 · Poster

ZEUS

Uses a training-free second-order predictor to reduce denoiser evaluations in diffusion and other ODE-based generative models.

Selected Highlights

ICML 2026 Spotlight

DecodeShare was selected for a Spotlight presentation.

HPCAsia ’25 Award

Scalable DCD received the Outstanding Paper Award.

Recent News

  • DecodeShare was selected as an ICML 2026 Spotlight.
  • ZEUS was accepted to ACM Multimedia 2026; I am a co-first author.
  • FlashSVD was presented as an AAAI 2026 poster.

View all news →

Earlier Milestones

Swipe or scroll horizontally to explore earlier updates.

Enrolled in Duke’s M.S. in Statistical Science program.

Received a scholarship from the Upsilon Pi Epsilon (UPE) Honor Society.

Placed 2nd in Deadlift and 3rd in Squat at the NCHSAA Triangle Novice.

Began a URECA research project with the Wake Forest IRSC Laboratory.

Last Modified: Thursday, July 30, 2026.