Kuartal Services

Own your intelligence.

Planned

Your institution's knowledge should make your AI better, not someone else's. Kuartal will fine-tune open-weight models on your own data, inside your walls, run the evaluations with you, and leave you with a model you own.

Data preparation inside the walls

The problem.

With per-token cloud AI, the institution rents intelligence and gets nothing durable back. With customer-owned open-weight models, the weights, the evaluation results and the improvement stay with the institution.

How it works.

  1. 01

    Data preparation inside the walls

    Customer data will be prepared and used for training on premises; it will not be sent offshore.

  2. 02

    Fine-tuning

    Kuartal will adapt open-weight models to the customer's language, documents and workflows, including Bahasa Indonesia.

  3. 03

    Evaluations

    Kuartal will run task-specific evaluations so the customer can see quality before going live.

  4. 04

    Ownership

    The aim is that the resulting model weights belong to the customer, under the terms of each agreement.

This capability is planned and not yet available.

What you get.

  1. 01

    A model adapted to your documents and workflows

  2. 02

    Evaluation results you can see before going live

  3. 03

    Training that happens inside your walls

Deployment options.

  1. 01

    Air-gapped on-prem

    Where data lives
    Inside your facility
    Internet connection
    None
    Who operates it
    Kuartal, through a controlled offline process
    Typical buyer
    Government, defence-adjacent, hospitals
  2. 02

    Connected on-prem

    Where data lives
    Inside your facility
    Internet connection
    Outbound only, under your policy
    Who operates it
    Kuartal
    Typical buyer
    Banks, universities
  3. 03

    In-country managed

    Where data lives
    Hosted by Kuartal in Indonesia, separate for each customer
    Internet connection
    Managed by Kuartal
    Who operates it
    Kuartal
    Typical buyer
    Smaller agencies, SMEs
  4. 04

    On-device

    Where data lives
    On the device itself
    Internet connection
    Not needed to run models
    Who operates it
    Your team
    Typical buyer
    Small offices, field teams

Compare deployment models

Questions.

Does any data leave our premises?

In an air-gapped deployment, no. In a connected deployment, outbound connections follow your own policy.

Who owns the model?

Kuartal deploys open-weight models, so the model runs and stays on your hardware. Ownership of model weights is set out in each agreement.

Which models do you use?

Capable open-weight language models that run without calling an external service. The list of supported model families will be published here.

Are you certified?

Not yet. Kuartal is an early-stage company and does not yet hold formal certifications. Our roadmap includes data-protection compliance work and security accreditation.

Where is Kuartal based?

Mataram, Lombok, West Nusa Tenggara, Indonesia.

Go deeper.

Let's talk.

Bring AI inside your walls.

Tell us about one workflow your institution wants to improve, and we will follow up.