I design and deploy automated workflows and tailored AI systems that catch every lead, clear the busywork, and hand you back the hours your business was eating.
Speed-to-Lead: scraped or inbound leads get scored 1-10 by Claude, answered in my brand voice, routed hot or nurture, and logged to a CRM before any send fires — built on n8n running on my own server, tested on my own leads first. Includes a live interactive demo.
Set before the first node was wired.
Vallar Browser Agent: an open-source Chrome extension that lives in your side panel, reads the page, and clicks, types, scrolls, and completes multi-step tasks autonomously — with per-site permissions, confirmation gates for sensitive actions, and a user-defined blocklist. Runs on any OpenAI-compatible endpoint, including fully local models via Ollama or LM Studio.
View on GitHubAutonomy with brakes, modeled on Claude for Chrome.
The 40-second Speed-to-Lead film here wasn't cut in an editor. An AI motion pipeline (HyperFrames) storyboarded it, built every frame as code in the handlit design system, and rendered it — using the real numbers from a live run: score 9/10, reply out in 7.4 seconds. The same pipeline produces launch films for client products. Music: "Static Motion" — Kevin MacLeod (incompetech.com), CC BY 4.0.
Open the Full Case StudyI design lightweight, self-testing n8n workflows that clean, route, and transform your data. Out of your head, into code.
Tailored AI agents trained to handle your brand voice, automate customer audits, and draft high-converting proposals on-brand.
A custom AI agent built around your business flows — sales, support, and ops handled by one operating core that runs your playbook, not a generic bot.
A two-minute form, read by me. It finds the waste loops and tells you honestly whether there's a build worth doing.
One workflow from your audit, built and shipped in two weeks. Small scope on purpose — you see it working before committing to anything bigger.
I become your automation department. A monthly partnership: new builds, monitoring, and a roadmap — instead of quoting every workflow separately.
The retainer, scaled to a department. I own your automation roadmap end to end — design, builds, monitoring, and the quarterly numbers — as a fractional operating partner.
If your contact page says "we reply within 24–48 hours," this is for you. I research your business the way a customer experiences it, then design and build the chat agent it actually needs — in your voice, with a clean hand-off to a human when it matters.
A plain-language report on how customers reach you today, where each path stalls, and which agent would close the gap. Useful even if you never hire me again.
Everything in the audit, then the build: a brand-matched agent on your website and your busiest channel, answering from your real content.
Your agent stays current after launch: prices, hours, policies, and the questions it couldn't answer last month — reviewed and fixed.
The research agent behind the audit is open source. Bring your own key, point it at any business, and it drafts the blueprint on your machine — nothing leaves it.
A free audit, no strings: tell me how your business runs in the short form here — two minutes, three fields — and I review it myself. No bots on this one.
Who it's not for: if inquiries are rare, or your week already runs without you, there's no build worth paying for yet — and the report will say exactly that instead of selling you one.
Start the Discovery AuditAll four are public repositories, with the run logs committed alongside the code. That's the proof on offer: read the code, run it on your own machine, check the numbers yourself.
A Chrome side-panel research agent. Point it at any business website; it studies the pages and drafts the full chat-agent blueprint — personality, the questions customers ask, scripts, and the numbers to watch. Bring your own key; runs on Ollama or any OpenAI-compatible endpoint, so nothing leaves your machine.
Get it on GitHub
Reads an inbox and sorts every message by what it actually needs — a reply, a task, or nothing — and drafts the replies worth sending. A logged run in the repo's evidence folder clocks 3.19 seconds.
Get it on GitHub
Drafts owner replies to customer reviews in the business's own voice, ready to approve and post. Three live executions logged in the repo, averaging 3.48 seconds each.
Get it on GitHub
Takes one finished piece of content and reshapes it for other channels without flattening the original voice. Logged runs ship in the repo's evidence folder.
Get it on GitHub