Private intelligence

Refad LLM

The language-intelligence layer underneath every product we build — deployable inside your own boundary, so the data that makes it useful never has to leave.

Discuss a deployment See the deployment models

The line the whole architecture exists to hold —Deploy the solution. Deploy the intelligence. Maintain control.

How it fits together

Products at the top. Intelligence at the core. Security underneath everything.

Our products are not isolated applications — they are layers of one connected ecosystem. What sits at the bottom is the part most vendors ask you to give up.

Owned AI products

REFAD · Matrix · SNext · Once

Enterprise AI solutions

Configured around one organisation’s use case, knowledge and security requirements.

Private AI infrastructure — Refad LLM

The foundation both layers above stand on.

Arabic & English Knowledge grounding & RAG Document understanding Assistants & agents Workflow automation Domain adaptation Model orchestration

Deployment

AI deployed around the organisation — not the other way around.

The model is chosen by data sensitivity, security requirements, integration needs and your own policy — not by what is convenient for us to host. All five run the same intelligence core.

Client on-premises

Both the solution and the intelligence layer run inside infrastructure you control. The recommended model for sensitive data.

Private cloud

A dedicated, isolated cloud environment rather than a shared tenancy.

Saudi-hosted

For data-residency and hosting requirements inside the Kingdom.

Refad managed

A managed architecture we operate on your behalf.

Hybrid

Internal systems, private data and approved external services under one controlled architecture.

Trust layer

Security and responsible AI are architectural decisions, not a policy page.

Security by design

Encryption in transit and at rest Role-based access control Tenant and data isolation Secure authentication & API access Audit logs and traceability Backup, recovery and resilience

AI safety

Model and output guardrails Human-in-the-loop workflows Prompt and input protection Output validation Monitoring and evaluation Hallucination-risk mitigation

Data governance

Purpose-based data use Access governance Retention controls Client-controlled knowledge sources Data-residency options Privacy-aware product design

Designed to support alignment with

NCA ECC NCA DCC NCA CCC Saudi PDPL SDAIA GenAI Guidelines

Designed to support alignment — not a certification claim. Where an assessment is required we will tell you what has and has not been assessed.

Deploy the intelligence, not just the software.

If your data cannot leave the building, that is the starting requirement rather than the obstacle. Tell us where it has to stay.

Talk to us