العربيةStart a project

Infrastructure

Our own servers, in Saudi Arabia.

The systems we build and the models inside them run on hardware we operate ourselves. We don't resell someone else's cloud.

Start a project
  • Dedicated servers for apps and databases
  • GPUs for reading, speech and search models
  • HPC management and job queues

Three kinds of machine

What runs on our floor.

  • Dedicated servers

    For client systems and their databases. The applications people log into run here.

  • GPUs for the models

    The models we keep for print, handwriting, speech and search are served on our own GPUs.

  • HPC and job queues

    We manage high-performance machines and how work is scheduled onto them.

Scheduling

Every job gets its turn.

Training, scoring and research jobs wait in a queue and run on the same machines that host the models, placed where there's room.

  • One queue for training, scoring and research
  • Jobs placed on free GPUs as they open up
  • Every job gets its turn
An illustration of the scheduler placing work on GPUs.

Apps and models together

Reading, listening and searching stay on the floor.

The systems people log into and the models beside them share the same machines. A page to read or a recording to transcribe is handled by a model on those machines, with no outside API involved.

Research compute

Clusters for research, managed.

The same work applies to research compute: we manage the machines, the queue, and how heavy jobs are scheduled, so researchers get time on the cluster without managing it themselves.

  • Cluster management
  • A queue and a scheduler for heavy jobs
  • Turns for every team, in the queue
An illustration of the scheduler placing work on GPUs.

Questions we hear

Is this a cloud service?

No. We run our own servers in Saudi Arabia. We don't resell someone else's cloud.

Where does our data live?

Client systems, the models beside them, and their data stay on our servers inside Saudi Arabia.

Can you manage our cluster?

Yes. We manage high-performance machines and the job queue that schedules work onto them, including clusters for research compute.

Works with

Need the system and the machines it runs on?

Tell us what you want to host, and how much work it does. We'll tell you how we'd run it.

Start a project