I design AI systems, and use them to help run projects.
Take this client request from intake to a reviewed change.
Clients are described, not named. References on request.
My workflow triages client requests on its own, completes the ones it can verify, and builds a daily preview theme for me to review. I maintain visibility into the overarching approach and I decide what gets merged.
A folder of plain Markdown notes that a person and an agent both write into, and a store underneath that turns those notes into something an agent searches at the start of a session. My system gives agents coherent ongoing memory, so they pick up the context and nuances of each project autonomously.
A system meant to answer clients’ questions about their own websites from three sources at once: our team’s handover documents and created outputs like design tokens, the platform’s help articles, and the website code, straight from GitHub and always up to date. Agents built it; I specified it, verified it, and ended it when it became more machinery than the problem deserved.
I put every connection between my AI assistants and outside systems behind one local gateway, Toolport. I didn’t build it. I chose it, configured it, and decided which of its protections were worth the friction.
For twelve years I ran digital projects for brands ranging from startups to Fortune 5. Now I design the systems that run the projects: what the agents do, what they’re not allowed to touch, and how their work gets checked before anyone sees it.
Download full resumeI’ve led delivery on large accounts including Coca-Cola, Principal Global Investors, and Purina.
I’ve managed Claude-powered builds of five large Shopify storefronts. I also built the agent pipeline that triages UAT feedback and fixes what it can, with a human reviewing every merge.
My one-person studio, for the client work I take on directly. It’s also where I try new tools first.
An agent may act where its work can be checked or undone, and has to stop and ask where it can’t.
That rule runs through the agent work here, from the read-only QA agents to the gateway that decides what an agent can reach. Drafting a client reply is encouraged. Sending it is mine.
My UAT workflow triages client feedback, completes the requests it can verify, and builds a daily preview theme. Nothing merges until I’ve reviewed it.
Read the case studyStanding rules for every agent, whatever its credentials would permit.
Claude today, GPT tomorrow, an open-weight model after that. I design systems that work with any of them: plain-text notes, clear rules, and one gateway for tools. They help whichever model is best for the job, and stay out of its way.
Claude, through Claude Code and the API, for most reasoning and code; OpenAI’s GPT-6 Astra and OpenRouter where they fit better. Agents run as a team in an issue tracker, each with a scope and a list of things it may not touch.
Tool connections behind one local gateway, plain Markdown notes as the source of truth, and a project-scoped memory layer that turns those notes into something an agent searches at the start of a session.
Deep in the Shopify Plus ecosystem: metaobjects, the Admin API, and the apps around them including Rebuy, Recharge, Klaviyo, and many others. I architect full site data models and run replatforms and store-to-store migrations end to end.
I improve efficiency by automating the tasks that must be done, so your operation is faster and more responsive. This isn’t saving money by laying off humans. It’s letting humans do more impactful work while AI handles everything else.
I’m open to projects, consulting engagements, and full-time roles. Tell me what you’re working on.