We're The AI Studio. We help mid-to-large companies put AI to work — not in slides, in production. Most of what we do runs through small builder pods that ship real systems for real clients on tight clocks.
At The AI Factory we work with world-class clients to deliver products, automations, POCs and agents that have direct impact on their business operations. Real stuff.
This page is a standing invitation. We don't run hiring rounds. We keep a bench of independent AI builders, and we add to it whenever we meet someone who fits. If that's you, the door is open.
The shape of the work
You ship AI products for our clients. The work is not (always) time-shaped — it's outcome-shaped. An engagement might be a 12-week build with a clear roadmap. It might be a one-night sprint to ship a prototype before a board meeting. It might be four hours of you orchestrating a swarm of agents to produce something a traditional team would have spent a month on.
You work in a micro-pod — usually solo, sometimes paired with one other builder, occasionally a trio when scope demands. You are supported by our AI Engineering Lead and our leadership; you are independent but not alone. You own the outcome — but not without support. We work with you to frame the right architecture, align on client expectations before you build, and make sure every engagement is scoped in a way you can actually ship clean.
The problem changes every engagement. One week it's a customer support agent on a messy enterprise knowledge base. The next it's a research automation for a PE portfolio company that has to be live before Monday. The next it's a real-time pricing tool for an insurer with a six-week runway. You don't pick the problem. You ship the solution.
We care deeply about excellence in execution and the quality of what we deliver — that's the whole product. We don't dictate the toolchain. You have full optionality: Claude Code, Cursor, agents, agent swarms, classical engineering, whatever combination produces the best result. The bar is the deliverable, not the method.
Who you are
Your CV doesn't matter much to us. Your last twelve months of output does. And the gaps between what you know and what an engagement requires should feel like things you'd close in a weekend, not in a quarter. We care about what you understand and how you think.
Signals that you've shipped
Some of these should be true:
- You've shipped at least one product where LLMs do real work — not a wrapper, not a demo, something with users or paying customers
- You have a GitHub, a personal site, or a feed full of weekend builds, half-finished experiments, or working prototypes that show how you think
- You've taken a vague brief, scoped it yourself, shipped it, and moved on — without anyone managing you through it
- You've written at least one piece of evals code that actually caught regressions in a production system, or you've felt the pain of not having one
- You can name three things current LLMs are bad at for engineers, with specifics, and three workarounds you actually use
Daily tools
Claude Code, Cursor, Open Code or equivalent. Some combination of Anthropic, OpenAI, and open-weights APIs. A modern shell setup you've actually customized.
Stack fluency
Python or TypeScript at production quality. Comfortable in at least one of: full-stack web (Next.js, FastAPI, similar), data pipelines, backend services. You know your way around vector databases (pgvector, Pinecone, Weaviate, or similar), orchestration frameworks (LangGraph, LlamaIndex, custom), and observability tools (Langfuse, LangSmith, Helicone, or you've rolled your own). What matters most is that you treat all of these as interchangeable — you can pick up anything.
Beyond the obvious
You've debugged a RAG system that returns confident garbage. You've built or worked alongside an agent that does multi-step tool use. You have a point of view on when to use fine-tuning, when to use RAG, when to just use a longer prompt, and when to question whether AI is the right answer at all.
Things you've shipped
Any of these:
- A customer-facing chatbot or agent that real users rely on
- A document or knowledge-base ingestion pipeline that handles messy enterprise data
- An internal tool that automates a workflow at least one team depends on
- A weekend hack with more than three users
- An OSS contribution to an AI tool you actually use
What we don't expect
That you've worked at a famous company. That you have a PhD. That you've used every tool in this list. That you fit a tidy seniority band. We expect that you can move fast, ship clean, and pick up new tools as the engagement requires. None of these are negatives — we just don't filter on them.
How we work — and how we pay
We try to focus on value, not hours. Our preference is to price every engagement on the outcome — what it's worth to the client — and pay our builders the same way. That alignment is what makes this model work for the people we want to attract.
Some clients buy in sprints. Some clients buy in hours. Some clients are ready for pure outcome-based pricing from day one. We meet clients where they are, and we work out the right structure with you per engagement. What we don't do is pretend that an hour of your time on a swarm-orchestrated build is worth the same as an hour of traditional engineering work — it isn't, and we won't price it that way.
We agree on the structure and fee before you start. We don't renegotiate mid-engagement except for genuine scope changes, handled in writing. This model only works for operators, not contractors. If you reach for a timesheet by reflex, this isn't the right fit. If being paid for what you produce rather than how long it took sounds like how work should always have worked, you're our person.
How we get to know each other
We aim to work with the highest-performing, high-agency builders in the market. People at that level don't audition. They don't do unpaid trials. They don't sit through five-round interviews. We don't ask them to. What we do instead is treat the early steps as a mutual filter — small, fast, designed to give both sides enough signal to decide whether the next step is worth taking.
A short take-home task. Something scoped, real, and reflective of the work we actually do. It should take you a few hours, no more. Or a few minutes and some tokens. We don't care. We're not looking for polish — we're looking at how you think, how you scope, and how you use your tools. You ship it however you want.
A conversation. We talk through what you built and why. We tell you about us, the work, and where the company is going. You ask the questions that matter to you. By the end of it, both sides have a strong read.
A real engagement. If there's mutual interest, we find a real engagement. The first paid project becomes the actual way we get to know each other — under real conditions, with a real client, with both sides paying attention. That's where we both learn whether this is a long-term fit. If it works, you're on the bench and we keep building together. If it doesn't, we part on good terms and you've earned a fair fee for real work.
No unpaid trials. No five-round panels. No ghosting. No live coding. Just two operators figuring out whether they want to work together — and if they do, doing real work to find out for sure.
What you get from this
A direct line to interesting problems. New domain, new stack, new client every couple of months. The variety alone is hard to find anywhere else. Great compensation and an insanely intense school for business and engineering.
A small, sharp network. No middle management. No sprint theater. No process for its own sake. The people you'll work alongside are picked for the same bar you're being held to.
Compensation that reflects what you actually produce. We've structured this whole model so that being fast and being good are rewarded directly, not arbitraged away by the agency in the middle.
Time between engagements that's actually yours. Project-based work means gaps. We treat that as a feature, not a bug. Build something. Travel. Rest. Come back hungry. Some of our builders have one or even multiple full-time jobs. Others are freelancers at the core. Many are founders.
When you're ready
Tell us what you've built recently that you're proud of — one paragraph or one link is enough — and anything else we should know.
We read everything. We respond to everyone we think might be a fit. If you don't hear back within two weeks, assume the timing isn't right — but feel free to come back later when you've shipped something new.
A few things people usually ask
Yes, as long as your availability for an engagement is what you say it is when you commit. We don't ask for exclusivity. We do ask for honesty about your capacity.
We're based in Amsterdam. Our clients are concentrated in Northern Europe but not exclusively. We work with builders anywhere in compatible time zones — Europe, UK, LATAM, occasionally further. The work is remote by default; we'll fly you to a client site when an engagement genuinely calls for it.
If it's our scoping miss, we make it right. If it's a genuine scope change from the client, we renegotiate in writing before continuing. If it's because you underestimated, we work it out — usually you eat some of it, we eat some of it, and we both learn.
Yes. We tell you what's coming, you tell us if it's a fit. We'd rather have you say no to three engagements and yes to the fourth than push you into something you're lukewarm about.
You work as an independent contractor through your own entity or as a freelancer. You handle your own taxes and insurance. We pay against agreed milestones, usually within 14 days of invoice.
Usually within a week of you reaching out. If we're slow, it's never because we're ignoring you — it's because we're heads-down on a delivery. We'll always come back.
That isn't currently an option, but it will be soon. Working with us as independent builders is by far the best way to stay ahead toward that opportunity. Business is booming and we're all into growth.