I build agentic systems that do real work.
Less AI theatre. More operating leverage. I design and ship AI operating systems for companies — Claude Code workflows, custom agents, memory and review loops — then train your team to run them without me. I’ve bootstrapped real businesses; I build AI the same way: scoped, shipped, reviewed.
hermes run "publish field note: copper"[plan] reading memory-bank … 3 sources[plan] plan written → 6 tasks[code] copper-18-year-problem.html written[code] research map wired · anchors checked[review] claims vs content-source.md … clean[review] contrast · reduced-motion · mobile … pass[deploy] vercel --prod … live✓ shipped — skill captured, the loop got smarter
Planned, written, built, and deployed by Hermes — my agent operating system built on Claude Code. The loop that shipped this page is the loop I install in client teams.
Co-founded Typo Coffee: bootstrapped to three profitable drive-thru locations, a roastery, 20+ wholesale customers, 25+ staff, approaching €2m annual revenue run-rate — now independently managed day-to-day.
Co-founded Sanity Cares during COVID: €2.5m revenue and 1.625m CE-marked face shields delivered to Ireland’s HSE in roughly six months.
Trinity College Dublin, Mechanical Engineering — First Class Honours, ranked 1st in class.
Every engagement starts with the real workflow, not the model. No pilots that die in a deck. Who hires me: founder-led companies and SMEs, funds and family offices that want practical adoption, and operators who tried chatbots and got nothing.
A map of your workflows ranked by where agents actually create leverage — control points and risks named before anything is built, a build-vs-buy view, and a prioritised 90-day plan you can execute with or without me.
The framework behind it is public → the agent adoption framework
Working systems in your stack — Claude Code workflows, custom agents, MCP connections into your tools, memory systems your agents can actually use, and human review gates on by default.
The pattern I run myself, installed for your team — working rules, a memory bank, plan and review loops, skill capture — plus two or three workflows redesigned end-to-end and hands-on training for the people who own them. You keep a playbook, not a dependency.
Ongoing ownership of your AI roadmap — new workflows each cycle, model and vendor decisions, systems kept current as models change, new staff trained.
Find where agents create real leverage — and where they don’t.
Scope the system: workflows, controls, data access, review gates.
Ship working agents in your tools, demoed weekly.
Your team runs the loop; I make myself unnecessary.
Skills and SOPs captured so the system keeps improving.
Everything is built in your stack. No platform, no lock-in, no licence I’m reselling.
Agents propose; named humans approve. Control points are designed in, not bolted on.
Every workflow is measured against time, throughput, and error rates — or it gets cut.
My own work runs on these exact systems. This site is the demo.
Hermes is my personal agent operating system, built on Claude Code: written working rules, plan-driven builds, coding and review loops, a private research memory bank with Python ingestion, verification before anything deploys, and skill capture so every cycle leaves the system smarter.
YouTube transcripts and research exports, ingested by Python scripts.
Obsidian-compatible vault; private → candidate → public lifecycle.
Claude Code: plans, subagents, skills, working rules.
Site pages and field notes, shipped from the same loop.
Verification checklist before anything goes live.
Repeatable patterns codified — the loop improves itself.
“Hermes isn’t a product. It’s Claude Code plus disciplined files. That’s the point — you can have one too.”
AI consulting from someone who has met payroll, suppliers, regulators, and rush hour. Systems I build are designed to survive contact with a real company.
Built from zero into three profitable drive-thru locations, a roastery, 20+ wholesale customers, and 25+ staff — now independently managed and my testbed for AI automation in a physical business.
CE-marked face shields delivered to Ireland’s HSE during COVID — €2.5m revenue, compliance, procurement, and subcontract manufacturing coordinated in roughly six months.
Six-storey Dublin city centre asset redeveloped into four apartments — project-managed across architecture, design, contractors, and delivery.
Full evidence map, including research and investing → Proof Index
Consulting is the day job; this is the research that feeds it. I invest privately across digital assets, AI, public technology markets, and commodities — early theme recognition across Bitcoin, Ethereum, Palantir, and Robinhood — and publish field notes tracing the AI buildout downstream: agents → data centres → power → uranium and copper.
Tell me what repeats, what it costs, and where it hurts. I’ll tell you honestly whether agents help — the audit exists to find that out cheaply.
Milan, Italy · remote across Europe · English native · learning Italian