space8 — Machine learning engineering

Q: Why do most machine learning proof of concepts never reach production?

Because a demo and a system are different things. A proof of concept proves the model can work; a production system needs architecture, data pipelines, integration into your stack, access control, monitoring, rollback, and a team that can operate it. space8 carries projects across that gap — the deliverable is a deployed system your engineers run, not a notebook.

Q: Can we use AI if our data cannot leave our infrastructure?

Yes. On-premise and in-perimeter AI is a solved problem — models can be developed, trained, and served entirely inside your infrastructure. space8 develops on your data, inside your perimeter, with access control, monitoring, and rollback as part of the deliverable.

Q: Should we build a custom AI system or buy an off-the-shelf product?

Buy when your problem is common and the product's assumptions fit your workflow. Build when your data is sensitive, your problem spans several disciplines, or no single product covers it. space8 recommends off-the-shelf components whenever they are the better answer, and you decide.

Q: How long does it take to build a production AI or ML system?

For a well-scoped problem, typically 2–4 months of build in two-week increments, usually after a short scoping phase. The deliverable is a deployed system, its documentation, and engineers on your side who have run a failure drill.

Q: How do we avoid becoming dependent on an AI vendor?

Demand it contractually and architecturally: documentation, handover, and a system your own engineers can operate without the vendor. space8 builds that way by default — ongoing support is available by the day, but never designed in as a dependency.

Q: What does it cost to develop a custom AI system?

It depends on the problem, the data, and the constraints — which is why space8 starts with a short scoping phase rather than a quote. You get a realistic plan, timeline, and budget before committing to the build.

Q: How do we know an AI system actually works?

Evaluation on your real data, not vendor benchmarks. Every system space8 delivers is evaluated on the client's own data, with the results documented — including where it needs human review.

Q: Where is space8 based and where does it work?

space8 is based in Lausanne, Switzerland, and works across Europe and beyond, remotely and on site. Most of its work is for institutions where data is confidential and errors carry consequences — finance, insurance, health, the public sector — but it takes on any problem that matters and needs a reliable ML system, from industrial operations to environmental monitoring. The team operates in English, German, and French.

space8

space8 — Machine learning, engineered to be relied upon.

space8 (Space 8 Sàrl) is an AI and machine learning engineering firm based in Lausanne, Switzerland. We design and build machine-learning systems for institutions where data is confidential, errors carry consequences, and reliability is non-negotiable.

What we do

Intelligent systems — bespoke systems engineered around your data and your constraints, from architecture through deployment, in your own infrastructure.
Machine learning development — model design, training, and evaluation, measured against the outcomes your organisation is accountable for.
Data & privacy engineering — encryption, anonymisation, and audit trails built into the system itself.
Architecture & advisory — an independent assessment of whether your data can support the objective, before a budget is committed.

From prototype to production

Each system is engineered around your data, your constraints, and your compliance perimeter. Where the data will not support the objective, we say so in the scoping phase, in writing. Typical engagements run 2–4 months in two-week increments; the deliverable is a deployed system, its documentation, and a team that has run a failure drill.