Hayden Gross
Helping companies and the people in them actually use AI.
As the Director of Technology at Solentrex, I took over a 14-repository platform, moved engineering in house, and cut engineering spend by about 95%. Prior to this I worked at Amazon as a logistics analyst, where I created reporting workflows that sped up escalations and raised productivity across the network. My career began with three years at Wedgewood Weddings running banquet operations, and some of my most valuable lessons came from there: how different teams work together, and how to make quick decisions when the day does not wait.
Projects and documents
The working environment
A place every AI session runs inside.
Memory that loads itself, rules the agents must follow, hooks that make it happen without me remembering, and verification before anything touches a customer or a dollar. It exists so nothing has to be explained twice and every session compounds on the last.
It is not only mine. The people with product context, not engineers, do the roof labeling that trains our model, and the website copy was written in house by the people who own the message; both work inside the same rulebook. The same pattern runs a second, smaller vault for a family business.
What it produced
Work
Each page has the same shape: what it was, what I did, what it produced.
Building the memory system from scratch
From 116 disconnected files to a five-layer memory that loads itself, saves itself, and never asks me to explain twice.
The brand kit and the content system
A plain file the AI reads before it writes: colors, voice, always and never. Behind the company's AI content since April.
Two AI tools, one brain, and a written cost policy
Two vendors' agents on one rulebook, a model and cost policy written after a 3.4 million token day, and non-engineers working in them.
Taking over a platform from an agency
What the environment produced first: engineering moved in house, fourteen repositories on one person, spend down about 95%.
Lender certification, and the money math
Software shipped inside the environment passed certification, and I built the checks that catch pricing and savings errors before customers see them.
Everything else
An in-house bill reader at about one cent per bill, a vendor decision proven the same day, an identity rebuild, ML work, integrations. The full list.
The same method, in a small business
A rental business runs on software I built and operate.
A family member manages about 50 rental properties. The app I built runs the business end of it: rent expected against rent received, leases, bills, tenant history, and collections flags, from a phone. AI does two jobs inside it, both fenced: it reads a messy spreadsheet and proposes rows a person approves one by one, and it answers questions over the ledger without ever inventing a number. Access is scoped by organization, vendor costs are capped in code, and every automated job writes to an audit table.
It is the enterprise rollout in miniature: identity, data governance, cost control, and a human in the loop, on a system a real business depends on every day. A household finance assistant with a local model that cannot see real account data, and a second memory vault, use the same approach.