Chaokun Chang Systems for AI
Menu

Research output

Publications

Research on agentic systems, LLM serving, video analytics, ML inference pipelines, and distributed training.

01

From LLM Inference to Agentic Workloads: Characterization and Implications for Serving Systems

Preprint

Chang, C., Zhou, Y., Lu, H., Fu, K., et al., Li, B., and Wang, W.

arXiv:2608.15127 · 2026

Paper

02

Crab: A Semantics-Aware Checkpoint/Restore Runtime for Agent Sandboxes

Preprint

Chang, C., Wu, T., Cao, L., Gao, W., and Wang, W.

arXiv:2604.28138 · 2026

Paper

03

Low-Latency, Resource Efficient Agentic Workflow Serving with FLARE

Manuscript

Fu, K., Chang, C., Shi, J., and Wang, W.

In submission · 2026

04

Hybrid LLM: Selective PD-Disaggregation for LLM Serving

Manuscript

Lu, H., Liu, M., Chang, C., Shi, J., and Wang, W.

In submission · 2026

05

Scalable Complex Event Processing on Video Streams

Conference

Han, C., Chang, C., Srijan, S., Lu, Y., and Lo, E.

ACM SIGMOD · 2025

DOI

06

Biathlon: Harnessing Model Resilience for Accelerating ML Inference Pipelines

Conference

Chang, C., Lo, E., and Ye, C.

PVLDB · 2024

PaperCode

07

Is Network the Bottleneck of Distributed Training?

Workshop

Zhang, Z., Chang, C., Lin, H., Wang, Y., Arora, R., and Jin, X.

ACM NetAI · 2020

PaperCode