Eighteen years of production software. Ten of them I ran my own company and grew it to a team of about fifteen engineers. Today I deliver the same scope with two or three people and an AI-assisted workflow. I've run it both ways, and measured it.
Available for contract and fixed-scope work — remote, across EU and US hours.
A civic map of small neglected places, live and open source.
No accounts at all: a phone takes a photo, a small model it carries judges clean, small or big, and neighbours confirm. The design problem is proving a place was improved when you know nothing about who says so.
Offline-firstOver 1,300 testsOwn photo model, 93 % clean or notVue 3 PWA, Cloudflare Workers, D1, R2
Nine sources swept nightly, scored by an LLM in two lanes, one ranked digest.
When the model misread a hard constraint four times, that decision moved out of the prompt and into code. Batching took scoring past 200 postings a night without adding a source.
Unattended200+ testsClaude API, Cloudflare Workers, KV
Each one marked by what I actually did. The colour is how close my hands were to the training.
trained by meshaped or prototypedpretrained, put to work
An agreed scope at an agreed price, working and deployed in days rather than months.
Models and automation inside a product people actually use, not a demo that impresses once.
Making the systems a business already runs on talk to each other, and stay in sync.
Senior engineering judgment on the days you need it, without a permanent hire.
A durable local queue, ordered replay, and conflicts resolved against server state on reconnect — so nobody has to think about being offline.
Isolated stages, each retrying on its own, with an audit record per step. One bad input never stalls the batch.
Pulls several internal APIs, narrows to the accounts that actually moved, and says plainly when nothing did.
Prices a set of items against a reference table and serves it — with the snapshot built free of personal data.
Small bridges pulling from one API and pushing to another: dedup, retries, and a mock of the far end so failures are testable.
2018 to now
I own the architecture and the product: 11 backend services and 6 front-end apps, multi-tenant across 27 locations in 9 countries. Offline-first field apps, AI document pipelines, computer vision, SSO and permissions, and the integrations that keep separate systems in sync. I write a large part of it myself.
2020 to now
Fixed-scope builds, AI integration and fractional tech-lead work, taken on directly.
2008 to 2018
Grew it from the first paying client to about fifteen engineers and more than a hundred delivered projects, owning architecture, hiring, delivery and the client relationships that paid for all three.
I talk to the people who will use the thing and agree the workflow before anyone writes code. Then a working prototype goes into real use as fast as possible, and everything after that is tuning against reality instead of guessing.
Or start with a one-week LLM pipeline review — $2,800 fixed
I usually reply the same day, and a 20-minute call is the usual start. Or read the one-page CV.