What we do
Most SMEs have a precise idea of how AI could help their business. What is missing is the team that turns that idea into something the business actually runs on. That is the whole job.
You get the next generation of AI-supported applications and workflows: software where the model does real work inside your business rather than sitting beside it as a feature. Requests are triaged as they arrive, documents are read and checked before anyone opens them, and the drafts, summaries and reports that used to take days come back in minutes — so your people spend their time on the exceptions instead of the routine. All of it built into the systems you already run, and costing what we said it would each month.
Building applications like that takes every discipline in one place. Kaimu is a one stop shop for AI-supported applications: business engineers, architects, ML and full-stack engineers, design and infrastructure, with decades of IT experience behind the AI. We also run our own: Lumosio and Librosio serve tens of thousands of users a day between them — so the questions that decide whether an AI system survives contact with real use are ones we answer for ourselves every month.
Working together
Whichever you pick, the work is the same: the same senior engineers, the same standard, the same refusal to ship anything we would not run ourselves. What changes is only how much of it sits with your team.
Prove it
SMEs with an AI idea and no proof yet.
We put your AI idea in front of reality: whether your data supports it, what a model can actually do with it, and what it will cost to run once it is live.
A workshop with the people who own the problem, a data readiness and running-cost assessment, and a working prototype you can click.
Start with a sprintAlongside you
SMEs with the plan and not enough hands.
Named AI specialists join your team, work in your repo and your standups, and leave the knowledge behind when they go.
Business engineers, architects and ML engineers by the day, working your process, your tooling, your board — scale up or stop on one month’s notice.
Pick a specialistLeave it to us
SMEs who want a working system, not headcount.
We answer for the result. Discovery through production — models, retrieval, interface and infrastructure — including the parts nobody demos: evaluation, monitoring and cost control.
Architecture and a delivery plan you can hold us to, a production application, tested and monitored, then handover to your team — or we keep running it.
Scope a buildThe expertise
The same three roles decide every AI project we have worked on. Whether the problem was framed properly, whether the system was designed for what it costs to run, and whether the models were ever held to evidence. Miss one and the project stalls somewhere nobody predicted.
Turns a fuzzy ambition into requirements the rest of the team can build against, and works out where AI helps and where plain software is cheaper.
Decides what the system is: which models, how data reaches them, where the boundaries sit, and what it will cost to run at your volume.
Pipelines, retrieval, tuning, and the evaluation that tells you whether any of it actually works.
Our products
Not client work. Both are ours, both are live, and both are paid for by the tens of thousands of people who use them every day. Every hard question about running AI at that volume — what it costs each month, what happens when the model is wrong, who picks up the phone — we answer for ourselves first.
ERP data → what actually moved your sales
Analytics has been a corporate privilege. The insight was always sitting in the order history, but reading it meant a data warehouse, a BI team and a programme no mid-sized business could justify — so the big players squeezed millions out of their numbers while everyone else ordered by feel.
Lumosio removes the price of entry: it connects to your ERP, enriches every record with the conditions around it — weather, markets, industry context — and runs regression analysis over the lot, showing which factors actually moved your sales and by how much.
The answer a retail group pays a department for, at a fraction of what it cost them.
No bias, no interpretation — just you and the page
Modern models have a scale problem: they have seen so much that they can no longer see your document inside it. Ask about one clause in one contract and the answer arrives assembled from ten thousand other contracts — well-written, certain of itself, and wrong in ways you will not catch unless you go and read the thing yourself. Which was the job you were trying to avoid.
Librosio does the opposite. It ingests your books, contracts and PDFs, digests them with a method of our own, and answers only from what is in them — every answer pointing at the passage it came from.
In legal mode it works the same way on a statute book: put your situation to it and you get what the law actually says about it — the article, the wording, the passage it sits in. Not what a model would advise you to do. It will not invent a precedent to be helpful, and it will not tell you what you want to hear.
No blending, no reconstruction, no polite invention. If the text does not say it, neither does Librosio.
Behind Kaimu
AI got easy to demonstrate and stayed hard to keep. A whole industry is growing around the easy half — workshops, maturity assessments, roadmaps that lead to more roadmaps — and leaving good companies holding a prototype nobody can run and an invoice nobody can explain. Kaimu was founded for the other half.
Co-founder · IT & AI Business Engineering
Thirty years of IT business engineering for blue-chip companies worldwide, through every wave of technology that was going to change everything.
After the Internet and mobile apps, ten years — already — of AI engineering, and mentoring in Applied Sciences alongside it.
Long enough inside large organisations to know which promises survive one, which stall in procurement, and which quietly disappear after the pilot. The difference was never the technology.
AI success is measured by fewer decisions at the same standard, not fewer people at a lower one.
Co-founder · IT & AI Architecture
Thirty years of IT architecture and development for Swiss blue-chip companies, across every stack that was going to be the last one anyone needed.
From client-server to web, mobile and cloud-native systems, with hands-on experience designing and integrating generative AI, LLM, RAG, agentic and document intelligence solutions.
Long enough building for real users to know which architectures bend, which ones break, and which ones get rewritten from scratch two years in.
AI can hold every fact the world has ever recorded. Information is the easy half; intelligence is what gets done with it.
