JP Howlett. This isn't a packaged product I'm reselling; it's the same
orchestration layer running my own portfolio of 20+ live projects right now, including the page
you're reading this on.
30+ Years Tech Experience.
Two modes: I build it for you, or I work alongside you until you can build it yourself.
Either way, the engagement ends with something real — working software, a fixed system, or a team that knows what it's doing.
I don't just show up to write code — I get in and understand how the business actually operates first.
When it's called for, that means process re-engineering, not just automation: finding where a
workflow is actually broken before deciding whether AI, custom software, or neither is the right fix.
Six of those years were on Wall Street, managing infrastructure, technical support operations, and call center ops —
the part of the job where systems break and someone has to own the outcome.
That background shapes how I approach AI work: I design for what happens when things go wrong, not just when they go right.
That infrastructure instinct goes back further — I ran a computer configuration center in the 90s,
building hardware and installing software by hand, and watched that whole discipline turn into automated
provisioning in real time. I was doing automated builds back then — systems provisioned automatically to a
specific user's spec, not just imaged off a master disk. Same problem I solve today with cloud instances and
Ansible; the tools changed, the discipline didn't.
After that, I ran my own bookkeeping company, building and running QuickBooks integrations for clients —
hands-on with the unglamorous plumbing that actually keeps a business's numbers straight, not just the theory of it.
I've also built for a federal government contractor: an ERP connector that took a week of manual work
down to about a minute of input and ten minutes of waiting. That's the kind of leverage I'm looking for on every
project — not a demo, an actual week given back.
My focus is control: how models behave in real environments, where workflows break, and how to design systems that keep humans in charge of outcomes.
This is not tool evangelism. It's practical leverage — architecture, implementation, and hard-won judgment — using the models available today.
I'm building the Promethean ecosystem (modpx.dev / moduspromethean.com) and AIGCSEP / The
Covenant, an AI governance protocol submitted to the IETF. The Emergency Stop function above and
the brokered-secrets architecture aren't borrowed concepts — they're the same governance model I
wrote the spec for, installed on your infrastructure.
