I build AI agent tooling, skill-based workflows, and creative AI integrations.
My work usually sits at the point where an AI agent stops being a chat box and starts becoming a useful operator: packaging context, running repeatable workflows, generating artifacts, previewing results, and checking quality before users have to clean things up manually.
| Project | What it is | Stack |
|---|---|---|
| Momoding | An open-source Android assistant that gets things done on your phone: reminders and events from one sentence, preferences it remembers, drafts that only go out when you tap Send, quiet routines, and experimental screen operation with approval cards. Built on OpenMinis. | Kotlin, Android |
| AgentBox | Portable packaging for AI agent environments: settings, skills, memory, session context, and project instructions bundled into a shareable .agentbox file. |
TypeScript, CLI |
| Hermes Knead WebUI | Hermes WebUI-based AI product shelf for creating, using, and shaping small AI products. | Python, WebUI |
| ComfyUI VIDU Nodes | Custom ComfyUI nodes for VIDU API workflows, including text-to-video, image-to-video, character-to-video, video upscale, and download helpers. | Python, ComfyUI |
Dokie AI PPT Skill
Built the original AI presentation skill workflow for generating, previewing, and quality-checking interactive HTML presentations with agents such as Claude Code, Cursor, and Codex.
Created during my time at MYZY-AI. I am no longer affiliated with the company.
- AI agent workflows that are repeatable, portable, and easy to inspect
- Tooling that turns model output into usable artifacts
- Creative AI systems for presentations, video, and interactive content
- Small interfaces that make complex workflows feel lighter
I am exploring agent skills, MCP-style tool ecosystems, and packaging systems that make AI workflows easier to share across projects, machines, and teams.


