On-premise setup
The Quote Engine works by reading your price list, cost table and technical drawings. These are a company's most sensitive assets, and most owners do not want them in the cloud, rightly so.
We are not asking you to upload your price list, your cost table and your technical drawings to the cloud. The server stays with you and prepares the quote where it sits.
The split goes into the contract. Your documents, index and audit log never leave the building under any condition. The only thing that can go out is a compute request, and you decide by policy where it goes.
you write the policy
Sending compute to a provider outside the building is a transfer of data abroad under article 9 of KVKK, Türkiye's personal data protection law, even if the provider keeps no logs and does no training. We do not build setups that claim "your data never leaves the building" while routing it to the cloud, and we write both planes down openly. If you prefer, the compute stays in too: we add a GPU node next to the box and the application side stays the same.
KVKK, Türkiye's data protection authority, published its Generative Artificial Intelligence Guide on 24 November 2025. Staff pasting customer lists into ChatGPT is now a named compliance question, usually first raised by your accountant or auditor.
In most companies the file server is either out of date or missing altogether. This server is both the document store and where the system runs, so part of that budget is usually planned already.
Your quote history, your specifications and your contracts are indexed on your own hardware. If you stop working with us tomorrow, the data stays with you, on your own machine.
The server at the entry tier has no discrete graphics card. Quantised small models do run on it, at a few tokens per second. That is why we separated the tiers, and we say so before the sale.
Indexing documents for AI search
Runs overnight as a batch job
Document classification and tagging
Short input, latency does not matter
Structured extraction from a document
Works with a queue: submit, then come back to it
Interactive quote drafting
A GPU node or remote compute is required
Your volume and the sensitivity of your data decide which tier you need. We measure both in the System X-ray.
Who it fits
A company for which document intelligence and background work are enough, with no need for speed on the quote side
Who it fits
A company that wants to run the Quote Engine interactively and keep its data in the building
Who it fits
A company that has to pass an audit, such as a defence, health or critical-infrastructure supplier
The price of the tiers is not written on this page. Once the scope is set, the quote comes out as a single number.
If the data sits on paper and in WhatsApp today, the directory gets built first. The server indexes a directory that already exists.
Someone inside is named to touch the hardware, or remote access is defined. That is how maintenance runs.
An on-premise server is chosen for data privacy and control. If cost is the only reason, we compare the cloud as well.
We do not guess the gain. The foreign-trade process map draws your operation in nine questions and marks the box eating the most time. Which line we discuss follows from there.