From prototype to reliable production
A demo that works on your laptop is not a product. This is the work that turns it into one. (Technically: MLOps — evaluations, monitoring, versioning, cost control.)
When companies call us
- The prototype works, but nobody dares put customers in front of it.
- After a prompt or model change, nobody can say whether it got better or worse.
- The token bill arrived and nobody can explain what it consists of.
- A contractor built it, left, and now nobody wants to touch it.
Most AI projects do not die because the model fails. They die in the space between “it looked good in the demo” and “it runs reliably for a thousand users a day and we know what it costs”.
This service is about exactly that space: deployment, monitoring, evaluations, prompt and model versioning, cost limits, and fallbacks for provider outages.
We do this for solutions we did not build too. If you have a half-finished AI project from someone else, we can take it over — the same way we take over half-finished apps.
How we work
1. Audit the current state
What runs, where, what it depends on, what happens on failure and what it actually costs. The output is a risk list ordered by impact.
2. Evaluations first
Until you can measure quality, optimising anything is pointless. We build a test set and automated scoring.
3. Deployment and monitoring
Kubernetes or serverless depending on load, request logging, alerts on error rate, latency and cost.
4. Hand over to your team
Documentation, an incident runbook and training. The goal is that you do not need us for every prompt change.
What you get
- A production deployment with quality, latency and cost monitoring.
- Automated evaluations that catch regressions before your customers do.
- Prompt and model versioning — you can always return to what worked.
- Cost limits and alerts so the invoice does not surprise you.
- An incident runbook and a trained internal team.
Indicative scope
- Audit of an existing AI solution
- €900 – €2,000
- Building evaluations and monitoring
- €2,500 – €6,000
- Production deployment and infrastructure
- from €5,000
- Operations and oversight
- from €350 / month
Prices are indicative and exclude VAT. The audit can be bought on its own — you are not obliged to continue with us.
Frequently asked questions
Will you take over a project someone else built?
Yes, we do it routinely. We start with an audit that says what can be salvaged and what is cheaper to rewrite. You get the result even if you decide to continue without us.
What exactly are these evaluations?
A set of inputs with expected outputs that runs automatically on every change. Without it, developing an AI solution is guesswork — you change a prompt, something improves and something else quietly breaks.
How can token costs be reduced?
Through caching, shorter context, routing simple requests to a smaller model and removing redundant calls. In practice a significant part of the bill can usually be saved without losing quality — but it has to be measurable first.
Does it have to run in the cloud?
No. We can deploy on your infrastructure as well. The deciding factors are data requirements, latency and who will operate it.
If you have a prototype you are afraid to release, this step is exactly what is missing.