Skip to content

Work with MX4

Selected engagements where our research applies.

Alongside our research, MX4 AI takes on selected consulting and engineering work for institutions and regulated organizations across MENA.

Discuss a project See the four areas
From constraints to architectureIllustration · not a delivered architecture
Where data may be processed
Connectivity
Languages in scope

In this configuration

Where the models run
On the institution’s own servers, inside its boundary.
What stays inside
Everything: documents, index, questions, answers and logs.
What is retrieved
Passages from permitted sources in Arabic, French and Arabic dialects, each with its section reference.
What is logged
Each question, the passages cited and the answer, in an audit log inside the boundary.
How updates arrive
By reviewed offline transfer; nothing connects out.

Change a constraint and the architecture recomposes. Illustrative only: each architecture is defined with the institution, for its own environment.

Areas

Where our work applies.

Each engagement addresses a defined technical need.

Sovereign AI architecture

Read about system constraints

Decide what to build, where it should run, and which controls it needs before committing to implementation.

Technical scope

Data residency, model selection, access controls and infrastructure constraints.

Example deliverables

  • Requirements and constraints assessment
  • Target architecture and technical choices
  • Integration and deployment plan

Private document AI

Example: searching internal knowledge

Give teams useful access to institutional knowledge through document search, source-grounded answers and tailored document workflows.

Technical scope

Document processing, retrieval, model integration and application development.

Example deliverables

  • Custom application and data pipelines
  • Connections to authorized sources and identities
  • Evaluation against representative tasks

Multilingual evaluation

Explore our multilingual approach

Evaluate and adapt AI for the languages your teams actually use: Modern Standard Arabic and its dialects, French, English and the terminology specific to your institution.

Technical scope

Multilingual documents, regional language variation, code-switching and domain-specific evaluation.

Example deliverables

  • Language and terminology requirements
  • Model and retrieval evaluation on your tasks
  • Workflow adaptations and documented limitations

Deployment in controlled environments

Read about private inference

Connect an AI system to your existing applications and make it operable in the environment your organization approves.

Technical scope

On-premise or private cloud, identity systems, network boundaries and operational controls.

Example deliverables

  • Application, data and identity integrations
  • Deployment configuration and validation
  • Operating guides and technical handover

The approaches behind the work

Multilingual evaluation, evidence-grounded retrieval, private inference and governed language data.

Explore our technology

The work in context

See the institutional workflows these services can support.

Explore use cases

Beyond implementation

Built for your teams to take forward.

Handover arrangements are defined in the agreed scope.

Decisions made explicit

Architecture choices, dependencies and control boundaries are documented so your technical teams can understand and evolve the system.

Evaluation in context

Tests reflect the agreed tasks, languages and operating conditions. Documented results and limitations inform system acceptance.

Operations documented

Configuration, deployment and operating procedures give your teams a practical basis for running the system.

Your project

What do you need to build?

Tell us the intended use, your environment and its constraints.

Discuss a project