On-Premise AI Services
We deploy private AI inside your infrastructure: an assistant on open-source models, search across your own documents, and a gateway that masks personal data before it reaches a public model.
Three services
Pick one or combine them. Each runs in your environment and under your rules.
On-Premise LLM
A ChatGPT-style assistant on open-source models, installed on your servers.
- Llama, Qwen, DeepSeek, Mistral and more
- Web chat with single sign-on
- OpenAI-compatible API for internal tools
On-Premise RAG
Ask questions across contracts, policies and reports, and get answers that link to the source.
- File shares, SharePoint and email
- Access rules per department
- Search across languages
Document Anonymization
Mask names, IDs and account numbers before text reaches a public LLM, and restore them in the answer.
- In front of OpenAI, Claude, Gemini or your own model
- Custom patterns for your own identifiers
- Same detection engine as Privacy Mask
When private AI makes sense
Data that cannot leave
Client files, internal reports and board papers that your policy keeps inside your own network.
Personal data in prompts
Teams that want public AI for drafting and analysis, without sending names, IDs or account numbers to it.
Answers from your own files
Knowledge spread across folders, mailboxes and systems that people search by hand today.
Isolated networks
Environments with no internet access, where a cloud AI service is not an option at all.
Any language
Documents and questions in the languages your team works in, Arabic included, with models tested on your material.
Your own integrations
Assistants that call your internal systems through an API, with your access rules applied.
How a project runs
Discovery
We look at your data, users, hardware and the rules your data falls under, and agree on the first use case.
Pilot
One team and one use case, tested on your own documents on your hardware or a test server.
Rollout
Sign-in, sources, access rules and monitoring for every team that needs access.
Support
Model updates, new sources and use cases, and training for your team.
Related products
Ready-made tools for the same problems.
Who uses it
Organizations that keep sensitive documents in-house.
Frequently asked questions
Have a question that is not here?
What should we prepare for the first discussion?
Bring one workflow to improve, examples of the material involved, the teams that need access and any rules about where data may go. Include your current hardware if private deployment is being considered.
Can we start with one team before rolling out company-wide?
Yes. The pilot focuses on one team and one use case. Use it to check the results and access requirements before planning a wider rollout.
What should our security team decide before the pilot?
Set the permitted data sources, who may access them and whether any public model calls are allowed. Agree these boundaries before choosing the service and connecting documents.
Can it run without internet access?
Yes. The on-premise LLM and RAG can run in an isolated network. Model files and updates are delivered in a way your security team approves.
How is this different from the Paperwork platform?
The platform is our product: bank statement analysis, fraud detection, document automation and the other tools, available over the API or deployed privately. These services are projects where we deploy general AI for your own team inside your environment.
Tell us about your data
A first call is enough to see which service fits and what the pilot would look like.