Frequently asked questions

It is computing infrastructure (servers equipped with graphics processors, storage, networking) installed on your premises or in a data center in your country, running your artificial intelligence models. Unlike online services, all processing happens in-house.

With an online service, your questions and documents are sent to the servers of a foreign company. With local infrastructure, the models run on your premises: no data leaves your network, the service works without Internet access, and it can be tailored to your documents and your business.

We favor latest-generation open models (including the Llama, Mistral, Qwen and Gemma families), selected according to your use cases, your languages and the applicable licenses. We compare several models on your own data before recommending the best fit.

French, English, Arabic and Portuguese are handled well by current models. For languages such as Swahili, Hausa, Wolof or Amharic, quality varies: we measure it on your real data and, where relevant, adapt the models using local corpora. We share the measured results with you, with no unverifiable promises.

No. The infrastructure is designed to run entirely offline. Updates can be delivered on encrypted, signed media. A link can be added for remote monitoring, only if you want one.

Not necessarily. A Compact Node can be installed in a fitted-out technical room. For larger configurations, we design the room or provide a sealed prefabricated module built to withstand heat and dust.

Every project includes a power assessment and an autonomy plan: UPS, batteries, generator and, where relevant, solar hybridization. As a last resort, systems shut down gracefully to protect the hardware and the data.

Yes. We connect the AI to your systems (document management, ERP, messaging, line-of-business databases, directories) through standard interfaces, respecting the access rights already in place.

No, by design. The infrastructure sends no telemetry outside and can operate fully isolated from the Internet. Authorized flows are defined with you and documented.

Non-disclosure agreements, access restricted to what is strictly necessary and logged, interventions escorted by your teams, no copies of data off-site. For sensitive projects, a dedicated intervention protocol is established.

Every project starts with an analysis of the applicable national framework (data protection law, sector-specific regulations, African Union instruments). This analysis is confirmed with local legal counsel when necessary.

Every project is priced individually: the cost depends on the number of users, the use cases, the site and the security requirements. We recommend starting with an assessment followed by a pilot, so that you invest on the basis of measured results.

Yes — in fact, we recommend it. A pilot on a priority use case validates the real value. The architecture is then expanded in phases, with nothing thrown away.

You do. The purchased hardware, the configurations, the adapted models and all data belong to you. Administrator access is handed over to you and reversibility is written into the contract.

Yes. We support government agencies in preparing technical submissions and respond to tenders, either on our own or in consortium with local integrators.

Yes, skills transfer is part of every project: training for administrators, technical leads and business users, with complete documentation in French and English.

We offer support contracts with written service levels: monitoring, preventive maintenance, critical spare parts stock, security updates and model upgrades.

Dubai is a global logistics hub, close to hardware manufacturers and distributors, with direct connections to many African capitals and a time difference of only 1 to 4 hours with the continent.

We consider projects across the entire continent. The feasibility of on-site work (logistics, regulations, security) is evaluated during the initial assessment, and we can partner with local firms.