AI Native Engineer
Frontend craft. Full-stack engineering. AI-powered experiences.

Blog   /   Notes   /   Juejin   /   Contact

I'm an engineer with a strong frontend foundation and a full-stack toolkit, building thoughtful interfaces, developer tools, and AI applications.

My work connects rich text editing, component infrastructure, backend services, and AI workflows. I focus on bringing model capabilities into useful interactions, connecting agents to tools and context, and making complex workflows maintainable.

Clear interactions. Traceable context. Results people can review and trust.

bilibili · Senior Software Engineer

June 2023 — Present

Build core web experiences, with a focus on content creation, reusable UI infrastructure, and web security engineering.

• Rich text editing: Led an editor architecture upgrade with Tiptap / ProseMirror, using a headless, plugin-based design. Improved content pasting across sources, formatting normalization, and custom content nodes.
• Component infrastructure: Built and maintained a component library from the ground up. Supported application modernization with Vue 3, TypeScript, a monorepo, and Vite, including automated testing and SSR compatibility.
• Web security: Owned SDK architecture and integration capabilities, focusing on runtime isolation, compatibility, and graceful degradation for reliable business integration.

Integrate LLM capabilities into product workflows, including AI text suggestions in collaborative editors. Use diff previews, per-suggestion accept / reject controls, and loading and error states to keep users in control of generated changes.

Build reusable context capabilities for coding agents through CLI tools, MCP servers, and plugin integrations. Work with Markdown + Git memory storage, local retrieval, and session-aware recall to support auditable memory and context reuse across repositories.

Develop applications on existing agent frameworks with LangGraph orchestration, specialized agent roles, LLM provider abstractions, and data adapters. Expose multi-stage workflows through CLI and API interfaces.

Work on the engineering around model calls: structured tool outputs, source attribution, explicit missing-data signals, asynchronous task states, regression tests, and deployment configuration. Keep outputs inspectable and failures easier to diagnose.

I share engineering ideas and lessons on my blog, in my technical notes, and on Juejin.

Email: minntaki@foxmail.com

_{Build with AI. Engineer with intent.
Stay curious. Turn ideas into working software.}