AI models and operations

A software company comes here after the first question is settled. Either the model you call today is not good enough on your own examples, or a system is already in production and nobody is watching it. One offer is a build. The other is care of something already live.

Private fine-tuned models start with an evaluation of the hosted model, not with a training run. We fine-tune an open model on your own data only when that evaluation says the hosted model falls short. If it is good enough, you stay on it. We run the model in your own cloud account, over two to six weeks, and leave an endpoint and a comparison report. It can start on its own or after an AI production sprint. Self-hosting is available on request.

Managed AI operations is the monthly agreement for a system that is already in production, whether we built it or we inherited it. We keep evaluations green, watch regression when a model or a prompt changes, and follow drift, latency and cost. Incidents get a response, and you get a monthly report. New features are not included. They are fixed-price change orders, agreed one by one. The monthly agreement does not fine-tune a model and it does not start from a blank page.

Choose the private model when the question is whether your own data should change the model, and where that model is allowed to run. Choose the monthly agreement when the system is already live and the question is who keeps it healthy. Contact us for pricing. The private model is a fixed price before we start. The monthly amount is agreed before the agreement starts, and we do not publish it here.