I am the one you would be working with.
Fifteen years of shipping software for people who needed one person accountable rather than a team to manage.
I started Consolidated Technologies in 2012 and have spent the years since doing the work most technology companies would rather not: migrating a city’s ERP data out of DB2, wiring Stripe Connect into a booking system that was never designed for it, building an SEO prioritisation tool for MLS inventory large enough to break the naive approach, standing up dashboards that executives actually opened.
None of that is glamorous, and all of it is where AI projects die. The models are the easy part now. The hard part is the system you are bolting them onto, the data underneath it, and knowing which of the twelve things you have been sold are worth building.
Before any of this I was a combat engineer in the U.S. Army, active duty and Guard from 2007 to 2015, responsible at peak for 54 people. That is where I learned that being the person accountable for the outcome is a different job from being the person with the best idea.
I built an AI product and kept AI out of the decisions.
AROK is a local-first system monitor. It watches a Windows machine for anomalous behaviour and explains what it found in plain language. The obvious build is to let a model decide what counts as anomalous.
I did not do that. Detection is deterministic and rule-based. The model narrates and assists, and it is never in the decision path. Security alerts have to be transparent and reproducible, and a probabilistic alert that cannot explain itself is worse than no alert, because people stop trusting it and then stop reading it.
That is the call you are hiring me to make. The source is public, so you can check that I actually made it.
Four things I will always do.
Tell you what not to build
Usually in the first two weeks. It is the least popular part of the job and reliably the most valuable, because the cheapest project is the one you cancel before it has staff on it.
Define done before we start
Written acceptance criteria and evals, agreed up front. If nobody can say what good looks like, nothing can be judged later, and the project ends in an argument rather than a launch.
Leave you able to fire me
Standard tooling, your repository, your infrastructure, documented as we go. A consultant who is hard to replace has optimised for the wrong thing.
Say it plainly when it is going badly
Early, in writing, with what I propose to do about it. You are paying for judgment, and judgment that only arrives as good news is not worth much.
Where this came from.
Worth a conversation?
Thirty minutes, no charge. If I am not the right person for what you are facing, I will say so on the call.