I'm Jacco van der Post, a freelance forward deployed engineer. I join your team, or build the whole thing on my own from A to Z: from an AI integration to a complete web application. And I stay until it runs safely in production and your people can run it without me.
01 // approach
AI labs and tech companies send forward deployed engineers to their biggest customers: engineers who sit inside the client's team until the thing actually works. The model works, but it's priced for enterprise budgets. I work the same way as a freelancer, for organisations that don't have an AI team of their own. No team to join? Then I build it myself, from first sketch to launch.
| Consultant | Outsourced project team | Forward deployed engineer | |
|---|---|---|---|
| Where | On the sidelines | In their own team, at arm's length | In your team, in your stack |
| Delivers | Advice and a slide deck | A project built to spec | Working code on your data |
| Domain knowledge | Generic | Second-hand, via briefs and specs | Learns your processes from the inside |
| Done when | The report is in | The spec is delivered | It runs in production and earns its keep |
| Afterwards | Implementation is your problem | Handover and a new statement of work | Your team runs it without me |
02 // services
Most organisations have run an AI pilot by now. Plenty never get past the demo, and the model is rarely to blame. Things stall where AI meets the systems you already run: permissions, data, integrations and upkeep. That's where I work.
ai.integration
Building LLMs (Claude, OpenAI or open source) into your application or workflow. Summarising, classifying, pulling data out of documents, or an assistant that actually knows your own material (RAG).
agents.mcp
MCP is the open standard that lets AI assistants and agents talk to your systems. With an MCP server you decide which actions are allowed, with permissions, logging and a human in the loop wherever it counts.
api.integrations
REST, GraphQL, webhooks and queues between CRM, ERP, payment providers, CMS and internal systems. With retries and monitoring, so it keeps working at three in the morning.
php.modernisation
Moving existing PHP applications to PHP 8.x, refactoring legacy code one step at a time and building new backends in Symfony or Laravel. No stop-the-world rewrite.
js.frontend
Front ends in TypeScript and Vue: dashboards, admin screens and real-time interfaces, including the chat and approval screens around an AI tool.
audit.advice
A fresh, independent look at your codebase, architecture or AI plan. You get a report with priorities: what needs doing now, what can wait and what you're better off leaving alone.
03 // demo
On its own, a language model knows nothing about your orders, customers, invoices or policies. It only becomes useful once it can query your systems through tools. Pick a scenario and look under the bonnet.
Simulation. In a real project this runs on your own data, with clear agreements on privacy, permissions and cost.
04 // process
No six-month commitment up front. Most projects start with one short, sharply defined step and a clear result. Then you decide whether we carry on.
1–2 days
We find out where AI or an integration genuinely saves time or money, or how healthy your codebase really is. You get a prioritised report and an honest estimate.
1–3 weeks
A working prototype on your data, in your environment. No demo theatre: you see straight away what it can do, what it costs and where it breaks.
per sprint or phase
Production-ready with tests, monitoring, documentation and a clean handover. Solo or side by side with your team.
block of hours or SLA
Ongoing development, maintenance and updates with agreed response times. You know exactly who to call, and I already know your system.
Logging, evals, fallbacks and cost controls are in from day one, not pencilled in for "phase two".
Designed with the GDPR and the EU AI Act in mind: as little data as possible, EU hosting or local models where it matters.
I plug into the systems you already run instead of rebuilding the lot. Faster, cheaper and far less risky.
My own company since 2005, several clients and my own kit. We agree on a result; how I get there is up to me.
05 // stack
I pick the right tool for the job, not the shiniest one. These are the ones I reach for every day.
06 // about
Business informatics taught me how organisations work, art school how people see, and thirty years in IT how to build software that stays up.
I studied Business Informatics at the HEAO in Arnhem and Fine Art at art school. I have worked in IT since 1993, built for the web since 1997 and run my own company since 2005.
My day job is large, business-critical web platforms for municipalities, hospitals, regional public health services (GGDs) and other public bodies: platforms where accessibility, security and continuity are not up for negotiation. I work on custom PHP, upgrades and migrations, integrations and performance, often embedded as the external developer in a digital agency's team.
What makes AI hard in production is familiar ground from that work: permissions, logging, privacy, accessibility and systems that must never go down. That's exactly the gap between a successful pilot and something people rely on every day. I use agents, MCP tools and automated code review in my own work every day.
"Having a programmer and a designer in one person is a real plus. It saves a lot of back-and-forth. Jacco is proactive in helping us find solutions."
07 // lab
I've been building games since 2011. First for iPhone and iPad, now in TypeScript with one codebase for web and iOS. Games are my test bench for performance, state and real-time architecture: 60 frames per second forgives nothing, and I bring that discipline to client work.
lab/arcade-shooter
A Galaga-style shooter with thirty enemy types, each with its own flight pattern, and waves that never play out the same way twice. The high scores come from a PHP API.
Phaser · TypeScript · Vite · PHP
lab/multiplayer-word-game
A rebuild of my own 2011 iOS word game, which already had real-time multiplayer, chat, push and social login. It now runs on a multiplayer server that shares its rules with the client, and it speaks several languages. Heading for the App Store.
Phaser 4 · TypeScript · Colyseus · zod
lab/turn-based-strategy
A turn-based strategy game for touchscreens. The game logic is a headless core with no render layer, fully tested and ready for web and mobile from a single codebase.
TypeScript · Phaser · Vitest
08 // faq
An engineer who doesn't advise from the sidelines but builds working software inside your team, on your systems, and stays until it delivers. Software companies have used the model with large customers for years, and AI labs now run FDE teams of their own. I work the same way as an independent: senior, hired for what you actually need, no enterprise contract required.
Yes. If you don't have developers of your own, I build the solution from A to Z: design, front end, back end, integrations, launch and the upkeep afterwards. You get one point of contact, and the code and documentation are yours.
That depends on the scope. I work by the hour or for a fixed price per phase. After an introductory call you get a proposal with a clear budget, so there are no surprises down the line.
Both. I'm based in Arnhem, in the east of the Netherlands, and prefer to work mostly remotely, with on-site sessions whenever they add something. Clients across the Netherlands, Belgium and Germany are welcome, in Dutch, English or German. Clients further afield are welcome too, remotely.
Yes. I've run my own company since 2005, work for several clients and use my own equipment. We agree on what gets delivered, not on how I spend my day. For Dutch clients, I'm happy to shape the assignment so it sits comfortably within the Dutch self-employment rules (Wet DBA).
Yes, and have done for years. I join a digital agency's team as the AI or PHP specialist, under your banner or mine.
Whichever fits the task, your budget and your privacy requirements best. That might be Claude, an OpenAI model or an open-source model on your own or European infrastructure. I build the integration so you can switch models later.
You do. After delivery you have the code, the documentation and the keys, and that's written into the contract. You're not tied to me, although I do hope you'll come back.
Yes. An integration that keeps falling over, an upgrade that keeps slipping, or an afternoon as a sounding board for your AI plans: small jobs are often how long partnerships begin.
09 // contact
Tick what you're after and tell me briefly what's in your way. I reply within a few working days. Prefer to skip the form? Email my inbox.