I design and build the wiring between business systems, so information moves between them without being retyped — and so the path it took can be reconstructed afterwards.
A practice management system, an accounting package, a document generator, a registry they can query, a mail server. What is usually missing is the wiring between them, so information gets carried across by hand — and the carrying is where the hours and the mistakes accumulate.
That is the work: not replacing what is already there, but connecting it into something that runs on its own and keeps a record of what it did.
I connect systems that were never designed to talk to each other: practice management, accounting, document generation, e-signing, public registries. The result is a workflow that runs end to end without re-entry. Most of the effort goes into the edges — what happens when a lookup returns nothing, when a document is rejected, when a counterparty answers three weeks late.
I aggregate public and third-party sources through their APIs — mapping data, company registers, published web content — into structured output an application can depend on. The difficulty is rarely the first request. It is rate limits, schemas that change without notice, and sources that go quiet, all of which have to be absorbed before anything downstream notices.
Where data residency or audit obligations rule out handing records to a third party, I deploy on private servers inside the EU. That extends to running language models on the same infrastructure instead of calling a hosted API, and to keeping logs in a form that can still be produced years later.
The compliance workflow is a reference implementation. The discovery application is work carried out for Netherstory.
A client-intake workflow built against Dutch Wwft obligations. It verifies a company against the national business register, screens the parties involved for sanctions exposure and politically exposed person status, classifies country risk, and produces a completed risk assessment together with a pre-filled engagement letter for a person to review and sign off.
Every lookup, score and decision is written to an audit trail retained for statutory record-keeping. It runs self-hosted against an EU-hosted language model, so client records never leave the environment. Built as a working reference implementation rather than for a paying client.
An application that turns any location into a personalised map of things worth finding. It collects candidates from public sources, categorises them, analyses the exterior and its surroundings from imagery, scores what it finds, and assembles the result around what the individual player is actually looking for.
It also improves itself. Community contributions are classified and placed on the map automatically, incoming reports are triaged without a moderator in the loop for the routine cases, and the verification photos players submit are fed back in to raise accuracy and engagement for whoever arrives next.
The constraint that shapes the design is coverage. It has to work in a town nobody has written a guidebook about, using only sources that are publicly available — which rules out curating anything by hand.
You know the architecture, what it costs to build and what it costs to run each month, before any code exists.
We walk through the process as it runs today and separate the parts that are genuinely mechanical from the parts that need a person. Not everything should be automated; the scope is whatever survives that conversation.
I write down the proposed system, the components it touches, and what it costs both to build and to run each month. You get this before any code exists, and it stands as a reference point rather than an opening position.
I build against your real data and your real exceptions, not a clean sample. Deployment is to your own instance, with the credentials issued to you rather than held on your behalf.
You receive documentation covering how the system works, how to change it, and how to operate it without me. Support continues afterwards on whatever terms suit the arrangement.
Systems run self-hosted where the requirement calls for it, and on EU infrastructure otherwise. Records are not routed through third-party services that the task does not actually require.
Built on open, portable tooling. If the arrangement ends you keep the system, the credentials and the documentation, and it carries on running.
Model size, hosting and storage are chosen against what the work actually demands. Over-provisioning is not a one-off decision — it is a bill that arrives every month for as long as the system exists.
Email is the reliable way to reach me. A short description of the process you want to change is more useful than a general enquiry.