Skip to content

On-premise edition

SenTra on your own servers

SenTra installed inside your network, with the translation model running on your own NVIDIA GPU server. Documents never leave the company, and the system works without an internet connection.

Request a pilot

What you get

  • Private deployment

    Docker-based installation on your GPU server, set up by our engineers remotely or on site.

  • Your terminology

    We load your glossaries and approved translations so the model uses your terms.

  • Integration

    REST API and web interface for staff; connects to document systems on request.

  • Support

    Updates, monitoring and a named engineer during the pilot.

Server requirements

Item Minimum Recommended
GPU 1 × NVIDIA GPU with 24 GB memory (e.g. RTX 3090, RTX 4090, L4) 1 × NVIDIA A100 or L40S, 48–80 GB
CPU and memory 8 cores, 32 GB RAM 16 cores, 64 GB RAM
Storage 200 GB SSD 500 GB NVMe
Software Ubuntu 22.04 or 24.04, Docker, NVIDIA driver and Container Toolkit Same
Network None after installation Internal network only

Security

  • No outbound connections.
  • Role-based access for staff.
  • Audit log of every translation.
  • Files deleted on a schedule you set.
  • Installed inside your perimeter, on your hardware.

How a pilot runs

  1. 1

    Call

    We agree on document types and languages.

  2. 2

    Installation

    Installation on your server takes 1–2 days.

  3. 3

    Two-week pilot

    Your team translates real documents.

  4. 4

    Review

    We review quality and speed together, then make a licence offer.

Request a pilot

Optional human review

For certified or notarised documents, a professional linguist can review the AI translation. Priced separately.

Also on the same private GPU infrastructure