Skip to content
View shuxiachai's full-sized avatar

Highlights

  • Pro

Block or report shuxiachai

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shuxiachai/README.md

Jingbo Chai · 柴靖博

AI Agent / RAG 开发者 · Open-source AI tools

我开发 AI 研究工具、带证据的报告生成应用和本地效率软件,并公开它们的实现与评测。
I build AI research tools, evidence-linked reporting workflows and local-first software.

悉尼大学计算机硕士 · Master of Computer Science @ The University of Sydney · 2027 届

邮箱:shuxiachai@163.com

精选作品 / Selected projects

ACA · 论文商业化评估助手

从研究主题或论文,生成带来源的商业化评估报告草稿。
Turn a research topic or paper into a cited commercialization report and scorecard.

Academic Commercialization Agent 历史版本的报告与评分界面

历史版本界面与报告样例,用于展示产品流程。

面向研究初筛:查看证据与可靠性提示,导出报告,并在任务中断后恢复执行。报告草稿仍需人工核查。

查看报告样例(历史版本) · 进入在线应用(需访问码或自备密钥) · 查看源码

BushfireReadyGPT · 山火准备报告助手

把澳洲地区与准备场景,整理成有证据、可审核的山火准备报告草稿。
A local-first bushfire preparedness workflow with hybrid RAG and human review.

BushfireReadyGPT 历史版本的本地报告创建界面

本地流程截图与演示视频,不代表当前云端版本的完整验收。

可查看资料与限制、审核草稿并导出文件。基于已注明来源的上游项目改造,用于作品展示与受控讨论,不用于实时火情、撤离或生命安全决策。

下载演示视频(89 秒 MP4) · 查看报告样例(v0.6.0) · 查看源码

地脉簿 LeyLineBook · 多账号任务管理工具

在电脑和手机上管理多个《原神》账号的日常与周期任务。
A local-first task manager with Windows desktop and offline PWA clients.

地脉簿多账号任务总览,展示三个合成演示账号的待办与完成情况

真实界面,使用合成演示账号与记录。

集中查看任务、记录完成情况,并通过备份在两端迁移数据。两端各自保存数据;工具不会连接或操作游戏本体。

打开网页版(PWA) · 下载 Windows 版 · 查看源码与使用说明

工程实现与评测 / Engineering details
  • ACA — Agent workflows & runtime reliability: deterministic retrieval freezes evidence before six LLM stages. Structured contracts, checkpoints and redacted tracing make failures inspectable. Recovery passed 30/30 offline fault-injection cases, within the published protocol and limits. Recovery evidence · Engineering case study.
  • Bushfire — Hybrid RAG & governed generation: combines deterministic analysis, retrieval, model generation and human review. The historical v0.6.0 structured-planning benchmark reports MRR 0.922 over 68 answerable queries and 5 safety negatives. Retrieval performance and report quality are evaluated separately. Retrieval evaluation · Upstream and attribution.
  • LeyLineBook — Product delivery & data reliability: backup migration, data validation, platform-specific credential boundaries, automated tests, Windows packaging and PWA deployment. CI.

Tools: Python · FastAPI · Pydantic · SQL/SQLite · JavaScript · Docker · GitHub Actions · OpenTelemetry.

交流与机会 / Opportunities

关注这里的 Agent 可靠性、RAG 评测与开源工具实践;使用问题与改进建议欢迎提交到对应项目的 Issues。

面向 2027 届 AI Agent、LLM 应用、Agent 平台与 AI 后端工程岗位
Open to 2027 new-grad opportunities in AI Agent, LLM Application, Agent Platform and AI Backend engineering.

Pinned Loading

  1. academic-commercialization-agent academic-commercialization-agent Public

    Turn papers and research topics into source-linked commercialization assessment drafts. Python, FastAPI and CrewAI; auditable scoring, checkpoint recovery and durable request receipts.

    Python 814 102

  2. BushfireReadyGPT BushfireReadyGPT Public

    Australian location + scenario to reviewable preparedness drafts. Hybrid RAG, evidence trails and human review. Local Ollama; password-protected cloud demo. Not emergency decisions.

    Python 2 1

  3. LeyLineBook LeyLineBook Public

    Local-first task manager for Genshin Impact players and studios, with Windows desktop and offline PWA clients.

    JavaScript 20 2