Overview

Chijun Sima

Efficient AI systems: scalable and cost-effective training and serving through system-algorithm co-design.

A concise profile of research direction, systems work, and representative publications.

Technical notes CV PDF
simachijun@gmail.com LinkedIn GitHub Scholar
Chijun Sima

Guangzhou, China

At a glance

Key facts and outcomes from my CV.

Quick facts

  • Senior Software Development Engineer, Tencent (WeChat), Jul 2020 - Present
  • OSDI 2022 co-first author for Ekko (low-latency model updates)
  • LLVM developer with commit access (Google Summer of Code 2018)
  • B.Eng. in Computer Science, SCUT, GPA 3.85/4.00, Rank 1/28

Research focus

  • Efficient AI (MLSys): scalable and cost-effective training and serving
  • System-algorithm co-design for production reliability and model freshness
  • Current focus: recommender and LLM systems under strict latency and cost constraints

Selected work

Condensed outcomes and project scopes across recommendation, data, and compiler systems.

Ekko

Low-latency model updates for large-scale recommendation

  • WAN bandwidth -92%
  • Machine cost -49%
  • 2.4s model-update latency

Data and feature platform

Safe, scalable training-data pipelines

  • WebAssembly in-process isolation
  • Locality-aware operator placement
  • Data movement reduced up to 1,200x

LLVM

Compiler infrastructure contributions

  • Semi-NCA and optimization pipeline improvements
  • Unified dominator-tree update APIs
  • Reported speedups up to 1980x

Selected publications

Two representative papers. Full list available on Google Scholar.

IEEE Access 2019

Dynamic Barycenter Averaging Kernel in RBF Networks for Time Series Classification

Kejian Shi, Hongyang Qin, Chijun Sima, Sen Li, Lifeng Shen, Qianli Ma

Further reading

Ekko architecture, deployment narrative, and supporting references are collected in a separate technical page.

Explore technical page Download CV