Independent operator / Agentic AI systems
Agentic AI systems. Built, operated, improved.
I turn complex, knowledge-heavy work into systems people can use. Then I keep them running.
Explore the work219 source notes / 14 localized editions
A traceable editorial pipeline, from source notes through localization, review, checkout, and versioned downloads.
A story becomes a finished animation.
Story → Characters → Scenes → Film
Turns a written story into characters, scenes, voices, motion, and a finished animated video.
One image. Seventeen ways to repeat.
Source ↗Explore all seventeen wallpaper symmetry groups. Processing stays on your device, with your source images in your browser.
Real work. Explore what went into it.Swipe to explore →
Inside the work / Build Your Home
A client’s construction experience became a traceable publishing operation: one source edition, fourteen languages, and a path from every requirement to the finished book.
See how it was builtSource tracking. Editorial review. Resumable localization. Payments. Versioned downloads.
Visit the live product02 / The method
First learn how the world works. Then build the means to change it. Let the consequences teach you where to act next.
Study the people, the workflow, the source material, and the constraints. Understand how the work behaves before deciding what to change.
Turn that understanding into a working system. Connect software, models, data, and tools around the change you want to make.
Put the system to work. Change the workflow, remove the bottleneck, and turn what you learned into a different outcome.
Monitor runs, investigate failures, control cost, and maintain the system. Learn from what changed, then use that understanding to act again.
03 / Work with me
I build and run agentic AI systems for founders, specialists, and teams with complex work to get done.
Map the work, choose a practical architecture, connect your data and tools, and build a working version with clear review and release criteria.
Design / Implementation / EvaluationManaged operation after launch: monitor runs, investigate failures, handle model and tool changes, manage cost, and keep the review process moving.
Monitoring / Maintenance / ImprovementImprove an existing workflow or prepare your team to own it. Clear documentation, practical training, and a handover grounded in how the system actually behaves.
Documentation / Training / HandoverResearch. Publishing. Document operations. Internal tools. Customer-facing products grounded in controlled source material.
If standard automation, search, or one model call is enough, that is where we start.
04 / Engineering notes and R&D
Follow via RSS ↗Architecture, decisions, failures, and the things still being figured out.
R&D / Local prototypePolyMind sends one brief to two different AI perspectives, then synthesizes the responses. Here is the saved run, what it cost, and what still needs testing.
Read the experiment ↗Loopy & Clippy is a local video editor for scenes, camera moves, timed text, and sound. Preview and export still diverge on some projects. The work continues.
Internal experiment / UnreleasedScope, process, case studies, and project fit.
05 / Start a conversation
Send me the workflow you want to improve. How it works today, who uses it, and what needs to change.
Let’s talk about it
