It runs.
The question is what it costs you.

You start with a diagnosis at a fixed price, not with a project.

AWS cost analysis: CHF 1'200 fixed, one to two days. If we find less in estimated annual savings than the fee, we do not invoice it.

AI visibility audit: CHF 1'800 fixed, three to five days. Ongoing monitoring afterwards is optional, CHF 900 to CHF 1'200 per month.

WordPress to AWS: CHF 4'500 fixed up to 15 pages, operation CHF 60 per month, content changes billed separately.

AWS Health Check: CHF 12'000 fixed, two weeks.

What grows out of it: customer portal from CHF 12'000 setup and CHF 1'500 retainer per month, SaaS in 8 weeks at CHF 35'000 fixed, back-office on autopilot quoted per case.

Migration and architecture mandates run at a day rate between CHF 1'200 and CHF 1'600.

Exact quote after the discovery call.

Yes. We take on subcontracts for AWS partner consultancies that need a senior architect on a specific mandate. K-I-Soft invoices your company; your client never sees the hand-off. We've worked this model before.

Switzerland, focus on DACH. Personal collaboration matters to us, so regular on-site exchange and a direct view of the running project are part of how we work. Remote where the project allows.

Quartiero and SCMC.ch are both live and public: Quartiero is our most recent end-to-end build, SCMC.ch has been in production since November 2025. The founder's full career history is on his LinkedIn profile. Happy to discuss specifics on a call.

German and English, fluently, both written and verbal. We've worked with teams across Switzerland, Germany, the US, and Hungary.

Yes, Claude and GitHub Copilot in daily work. But every line is reviewed, understood, and owned by a senior architect with 30 years of engineering experience. That's what catches the mistakes AI still makes. That's what you're paying for, not the typing.

The AI components in the back-office and SaaS solutions run through Amazon Bedrock inside your own AWS environment. That has concrete consequences for your data. Your inputs and the model's responses are not used for training, not shared with the model providers, and do not leave the AWS network. Everything is encrypted in transit and at rest, and on request with your own key. Processing stays in the chosen AWS region, for Swiss projects usually Zurich. That keeps FADP- and GDPR-compliant processing possible, which we have already implemented at SCMC, a production Swiss cybersecurity platform.

Yes, as a standalone two-week audit, fixed price, separate from the build-then-retain solutions above. Architecture review against the Well-Architected pillars (cost, security, reliability, operational excellence, performance), written report, prioritized findings, 90-day roadmap. Often pays for itself in the first month from cost savings. CHF 12'000 fixed. Ask on the call if this fits.

More and more people no longer search at Google but ask ChatGPT, Claude or Perplexity. These systems read websites differently than a human does, they see only what is in the delivered document directly, and much of what is assembled in the browser stays invisible to them. We build sites so their content is readable for these systems from the first request, and we make visible on a dashboard which AI crawlers actually fetch the site. More on this in the post: What does AI-findable mean?

K-I-Soft, an independent Swiss software practice, senior-led by Istvan Kappelmayer. We build digital foundations serverless on AWS: the website is the entry point, and the same foundation later carries your customer login, internal workflows and automation. The data sits in Switzerland, and you work directly with the architect who also builds. References are Quartiero and SCMC.ch, both live and public.

By getting the prerequisites right. AI systems read only what is in the delivered document directly, and they need clear, structured facts about your company. We build sites so their content is readable from the first request, add structured data, and make visible on a dashboard which AI crawlers actually fetch the page. We create the prerequisite and make it verifiable. We cannot guarantee a mention, that is ultimately decided by each system.

For a pure business-card site, a web host is often enough. But as soon as a customer login, your own workflows or automation come into play, a foundation that grows with you and only costs when used pays off. That is exactly what serverless on AWS is built for. If your plan stays small, we will tell you so honestly in the first call.

It depends on the scope. An AWS cost analysis takes one to two days, an AI visibility audit three to five days, a WordPress migration two to three weeks, an AWS Health Check two weeks. A SaaS foundation we deliver as a fixed-price package in eight weeks, and migration mandates run longer and in stages. After the first call you get a concrete timeline for your case.

Classic Java on Lambda has a slow cold start, often a second or more, because the JVM has to spin up first. Quarkus Native, compiled with GraalVM to a native binary, starts in double-digit milliseconds. For a serverless architecture where functions start on demand, that is the difference between a noticeable delay and a response that feels instant. At the same time memory usage stays low, which lowers the cost.

DynamoDB is serverless in the same sense as the rest of the architecture: no instance running and being maintained, no fixed cost at zero load, automatic scaling under spikes. For most SME and SaaS applications a well-modelled single-table design is fully sufficient, and faster and cheaper than a permanently running SQL instance. Where real relational queries are needed, Aurora Serverless comes in addition, not instead.

Infrastructure as code with AWS CDK, so no hand-clicked resources, everything versioned and reproducible. Delivery via GitHub Actions with end-to-end tests. Observability is built in from the start, not bolted on later. On retainer engagements we take over operations, monitoring and on-call. You maintain nothing yourself.