Open-source, zero-dependency tooling for the full lifecycle of medical-device AI, organised as a plugin matrix on LGD — a declarative, industry-neutral skeleton in which every capability is an independently loadable module. Medical devices are the reference domain; the skeleton itself is industry-neutral and extensible to AI agents, cybersecurity and other verticals.
Focus: open-source tooling for medical-device regulatory affairs & AI governance, published as an individual project by @stevenzhao07. Pure technical input, no commercial promotion.
- medical-ai-toolkit — the single navigation portal. It auto-discovers and routes all modules; start here.
Repository count verified via GitHub API on 2026-10-16: 26 public repositories under
@stevenzhao07.
| Category | Repositories |
|---|---|
| Portal / profile | medical-ai-toolkit · stevenzhao07 (this repo) |
| AI governance & trust methodology | ai-a3-laws · ai-governance-checklist · ai-model-passport · delivery-guard |
| Medical-device lifecycle (QMS / RA) | design-controls-toolkit · device-regulatory-map · iec62304-software · iso14971-risk-file · manufacturing-gmp · clinical-evaluation · biocompatibility-iso10993 · sterilization-validation · pms-vigilance · udi-assignment · cybersecurity-medical |
| AI / ML tooling | ml-validation-toolkit · samd-ra-toolkit |
| Interop / data / labeling | fhir-ortho-toolkit · omop-ortho-toolkit · multilingual-ifu |
| Cross-domain / ecosystem | silent-failure-catalog · uibc-core · awesome-healthcare-agents · machine-sync |
- Read the Contributing Guide (hosted in the portal).
- Pick a task from Good First Issues if you are new.
- Open an Issue or PR against the relevant repository. Every change goes through a branch + pull request; nothing is merged to
mainwithout review. - Report security or regulatory-accuracy concerns via the portal's Security policy.
We separate four distinct levels and never conflate them:
| Tier | Meaning | Current state (2026-10-16) |
|---|---|---|
| 🟢 CI | Automated checks pass in GitHub Actions | medical-ai-toolkit harness green (last success 2026-10-13 02:55Z) |
| 🟡 Functional self-test | Module ships a runnable self-test | Partial — most CLI modules include a self-test path; coverage varies per module |
| 🔵 Expert review | Domain-expert sign-off | Pending — these are reference implementations, not clinically validated |
| ⚪ Official source verification | Regulatory constants traced to a dated official source | Partial — where annotated (e.g. GMP 2025 #107 verified 2026-10-12); others marked "verify before use" |
⚠️ Nothing here is a medical device, a certified regulatory submission, or clinical advice. Tools output decision-support references; users remain responsible for confirming against current official texts.
Declared intent: code primarily Apache-2.0, documentation/content under CC BY 4.0 (some reference texts CC0), where a LICENSE file is present. Across the account, licenses are mixed per repository (Apache-2.0 / CC BY 4.0 / CC0 / MIT) — see each repo's LICENSE. A LICENSE file for this profile repo is pending owner approval (deferred); until filed, the license here is proposed, not yet final.
Some repositories originate from or are mirrored with the associated account @zhaoxinghua09-cell (e.g. uibc-core, silent-failure-catalog, awesome-healthcare-agents). They are listed here for discoverability but their upstream ownership is noted in each repo's own README.
This account and its documentation were produced with AI assistance. The specification for the account reorganisation was drafted with ChatGPT and reviewed by the account owner; execution was performed by a WorkBuddy agent on the owner's machine. AI is assistive only and does not replace domain-expert review or official-source verification.
Maintained by @stevenzhao07 · Last layout update: 2026-10-16.
面向医疗器械 AI 全生命周期的开源、零依赖工具集，以 LGD（一种声明式、行业中立的骨架，其中每项能力都是可独立加载的模块）上的插件矩阵形式组织。医疗器械是参考领域；骨架本身行业中立，可扩展至 AI 智能体、网络安全等其他垂直领域。
定位： 医疗器械法规事务与 AI 治理的开源工具，由 @stevenzhao07 作为个人项目发布。纯技术性输入，不涉及商业推广。
- medical-ai-toolkit —— 统一导航门户，自动发现并路由所有模块；从这里开始。
仓库数量经 GitHub API 于 2026-10-16 核验：
@stevenzhao07名下 26 个公开仓库。
| 分类 | 仓库 |
|---|---|
| 门户 / 主页 | medical-ai-toolkit · stevenzhao07（本仓） |
| AI 治理与信任方法论 | ai-a3-laws · ai-governance-checklist · ai-model-passport · delivery-guard |
| 医疗器械生命周期（QMS / RA） | design-controls-toolkit · device-regulatory-map · iec62304-software · iso14971-risk-file · manufacturing-gmp · clinical-evaluation · biocompatibility-iso10993 · sterilization-validation · pms-vigilance · udi-assignment · cybersecurity-medical |
| AI / ML 工具 | ml-validation-toolkit · samd-ra-toolkit |
| 互操作 / 数据 / 标识 | fhir-ortho-toolkit · omop-ortho-toolkit · multilingual-ifu |
| 跨领域 / 生态 | silent-failure-catalog · uibc-core · awesome-healthcare-agents · machine-sync |
- 阅读**贡献指南**（托管于门户仓）。
- 新手可挑选**适合新手的议题**。
- 在对应仓库提出 Issue 或 PR。所有变更都经由分支 + 拉取请求流程；未经评审不得合并到
main。 - 安全或法规准确性问题请通过门户的**安全策略** 报告。
我们区分四个独立层级，绝不混为一谈：
| 层级 | 含义 | 当前状态（2026-10-16） |
|---|---|---|
| 🟢 CI | GitHub Actions 自动化检查通过 | medical-ai-toolkit 测试框架为绿色（最近成功 2026-10-13 02:55Z） |
| 🟡 功能自测 | 模块自带可运行自测 | 部分 —— 多数 CLI 模块含自测路径；覆盖随模块而异 |
| 🔵 专家评审 | 领域专家签核 | 待定 —— 这些是参考实现，未经临床验证 |
| ⚪ 官方来源核验 | 法规常量追溯到带日期的官方来源 | 部分 —— 标注处（如 GMP 2025 #107 于 2026-10-12 核验）；其余标记「使用前请核验」 |
⚠️ 本账号下没有任何内容是医疗器械、经认证的法规提交文件或临床建议。 工具输出决策支持参考；使用者仍须对照现行官方文本自行确认。
声明意图：代码主要为 Apache-2.0，文档/内容为 CC BY 4.0（部分参考文本为 CC0），以各仓实际存在 LICENSE 文件为准。账号内许可证按仓库混合（Apache-2.0 / CC BY 4.0 / CC0 / MIT）——见各仓 LICENSE。主页仓的 LICENSE 文件待所有者批准（已延期）；正式归档前，此处许可证为提议状态，尚未最终确定。
部分仓库源自或镜像自关联账号 @zhaoxinghua09-cell（如 uibc-core、silent-failure-catalog、awesome-healthcare-agents）。此处列出仅为便于发现，其上游归属在各自仓库 README 中注明。
本账号及其文档由 AI 协助完成。账号重组规范由 ChatGPT 起草并经账号所有者审阅；执行由所有者机器上的 WorkBuddy 智能体完成。AI 仅为辅助，不替代领域专家评审或官方来源核验。
由 @stevenzhao07 维护 · 最近布局更新：2026-10-16。