🔒 Sovereign AI · ISO/IEC 27001 · SOC 2 · NIST standards

The full power of AI.
Without ever letting your data leave your walls.

AfricAIsoft builds AI agents and digital platforms that put the tools of large institutions within everyone's reach, from banks to independent traders. Your data stays with you. Period.

Who are you? Pick your profile and we'll show you the right solution in 30 seconds.

✓ Free 30-min assessment ✓ Free process mapping ✓ No commitment
🔒 100% local execution
⚖️ Compliant with Law 25 · ECOWAS · Malabo
🌍 French & English, jargon-free
✋ Humans always decide
The real risk

Every prompt sent to a foreign cloud
is a copy of your data beyond your control.

💧

Silent leakage

Confidential documents, client files, trade secrets: once they reach a third-party provider, you no longer know who reads them or where they are copied.

⚖️

Regulators don't forgive

Law 25 in Québec, national laws in West Africa: one piece of sensitive data processed outside the jurisdiction means fines and lost trust.

🔗

Costly dependency

Per-token billing that explodes, rising prices, services that change overnight. Your AI belongs to someone else.

There is another way: artificial intelligence that stays with you. All the productivity of a modern AI assistant, without ever giving up control of what matters.

Flagship offer · Banks · Firms · Governments

Sovereign AI: the full power of large models, without your data ever leaving.

An inference appliance installed on your premises runs AI models locally. Your confidential data goes through no cloud, feeds no third-party model and stays under your exclusive control, while keeping the productivity of a modern AI assistant.

🔒

Total confidentiality

Data physically stays on your premises. No transit, no copy at a provider, no external exfiltration surface.

⚖️

Regulatory compliance

ECOWAS Supplementary Act, Malabo Convention and national laws (CIL, ARTCI, CDP, NDPC, Law 25…) respected by design.

💸

Controlled costs

One-time appliance purchase: no usage billing, no per-token cost, no cloud subscription that explodes with volume.

📡

Offline / air-gap

Works without Internet. Ideal for isolated networks, remote sites or areas with weak or unreliable connectivity.

🎯

Tailor-made

Fine-tuning on your own business data and processes, for an assistant that speaks your organization's language.

🛡️

Independence

No vendor lock-in or geopolitical dependency: you control the model, the updates and the availability.

Step 1 Free

Google Meet interview · 30 min

Qualification of your need, regulatory context and priority use case. No commitment.

Step 2 Free & included

BPMN 2.0 mapping of one process

Free modeling of one business process requiring sovereign AI, delivered before any invoicing.

Three support tiers

Pooled on-site FIFO deployment

Tier 1 · Group
Essential
Group tour

Entrepreneurs & independent consultants comfortable with local inference (e.g. IT developers familiar with LM Studio).

  • Supercomputer on a bastioned network (home / small business)
  • Installation of an inference tool (LM Studio, Unsloth Studio)
  • Security governance configurations
  • 1-hour training dedicated to the entrepreneur
10 seats · group cohort
€4,170
fees €3,330 + FIFO €840 · /client · appliance excluded
Book my assessment →
Tier 2 · Express Group
Enterprise
Express tour

SMEs & organizations with an in-house IT team wanting a supervised, accelerated deployment.

  • Supercomputer on a bastioned network (corporate network)
  • Security governance configurations
  • Installation of an inference tool (LM Studio, Unsloth Studio)
  • 1-hour training for IT administrators & the CTO
10 seats · express cohort
€5,280
fees €4,440 + FIFO €840 · /client · appliance excluded
Book my assessment →

Hardware catalog local inference appliances

x86 AI APU · unified memory · 100% local execution · up to 235B parameters
Class 1 · Compact node
Compact appliance
16 cores · 64 GB LPDDR5X · 1 TB NVMe · ~50 TOPS NPU · models ≤ 32B
€1,404indicative
Class 2 · Best balanced
Versatile appliance
16 cores · 96 GB unified · 2 TB NVMe · VRAM up to 96 GB · DeepSeek 70B / Qwen3
€1,773indicative
Class 3 · AI hub & cluster
Cluster appliance
16 cores · 128 GB unified · dual 10 GbE · server mode · models up to 235B
€2,069indicative

Indicative hardware prices (1 CAD ≈ €0.616, June 2026), excluding taxes, shipping and FIFO installation. The supercomputer price is excluded from the three support tiers and agreed with the client based on needs and stock availability. Hardware examples: Abacus Go1, ASUS Ascent GX10. No brand is affiliated with AfricAIsoft.

Book your 30-min assessment.
Free, no commitment.

I help banks, accounting firms and governments deploy 100% local AI infrastructure where data sovereignty is non-negotiable. In 30 minutes, we qualify your need and you leave with a concrete plan.

  • 1
    30-min Meet interview. Qualification of your need, regulatory context and priority use case.
  • 2
    BPMN 2.0 mapping of one of your processes, modeled for free before any invoicing.
  • 3
    Tailored proposal: support tier, choice of open-weight model and appliance.
🎁 30-min interview + BPMN 2.0 mapping of one process, free and included in every tier.

Meetings can be held in English or French.

Step 1 / 3
📅 Choose a date

Mon, Fri & Sat · 9 AM to 7 PM (Montréal time, ET) · 24h minimum notice

Step 2 / 3
⏰ Choose a time slot
Loading availability…
Step 3 / 3
📝 Your details
✅

Meeting confirmed!

A Google Calendar invitation has just been sent to your email address. You'll find it in your inbox.

Any questions in the meantime? consulting@africaisoft.africa

The rest of the ecosystem

Our other AI products, for individuals & builders

Agentic AI trading
● Available
LLM Council Expert Trader
Multi-agent EA · MetaTrader 5 · Forex / Crypto / Indices

Expert Advisor for MT5 powered by a multi-agent council (HTF, Structure, Strategy, Execution). Hybrid routing between LM Studio (local), Together.ai and OpenRouter. LoRA and fine-tuned model support.

  • 4 AI agents (SMC/ICT)
  • Local inference via LM Studio
  • LoRA / fine-tuning support
  • Sold on MQL5 Market
View on MQL5 Market →
Tools for quants & AI developers
⚙ Gumroad tool
GGUF Fusion Pro™
LoRA + GGUF model merging · for LM Studio

A tool to merge your LoRA adapters with a base model to produce a GGUF file usable directly in LM Studio. Ideal for deploying your fine-tuned models locally.

  • LoRA → GGUF merge
  • LM Studio compatible
  • For local, off-cloud use
  • Available on Gumroad
View on Gumroad →
⚙ Gumroad tool
Fine-Tuning JSONL Generator
Training dataset generation for trading LLMs

Generator of ready-to-use JSONL datasets to train your own trading LLM agent on Together.ai or any other fine-tuning platform. Optimized for SMC/ICT concepts.

  • Together.ai-compatible JSONL format
  • SMC/ICT templates included
  • For quants and researchers
  • Available on Gumroad
View on Gumroad →
Applications & platforms
○ Coming soon
AI Delphi Council
Multi-model decision council · cloud productivity tool

Several large language models answer, review each other and rank each other blindly, then a chair model writes the final synthesis with its agreements, disagreements and caveats. Drop in your documents and images as context, export the result to Word, PDF, PowerPoint or spreadsheet.

  • Multi-model deliberation (Delphi method)
  • Your documents & images as context
  • Built-in web & deep research
  • Word / PDF / PowerPoint / spreadsheet exports
  • Prepaid credits, no subscription · French-first
See the demo →
Join the access list →
💬 WhatsApp chatbot
DubaPass
Digital twin of your property title · via WhatsApp

A conversational chatbot app on WhatsApp that lets owners create a secure digital twin of their property title without complicated forms, in a few natural exchanges. Designed for the peace of mind of property-owning families.

  • Conversational via WhatsApp
  • AI-guided digitization
  • Digital twin of the land title
  • For French-speaking African owners
Learn more →

Find all our products on MQL5 Market, Gumroad and LinkedIn AfricAIsoft.

Strategic model · AfricAIsoft Research

TOMBOUCTOU

A RAM-first language model architecture with compressed semantic memory, powered by the SANKORÉ reasoning engine.

Current version
Tombouctou-1-moe_ijazahSankoré1
Whitepaper summary

The problem. Modern language models are prisoners of GPU VRAM: a few global companies own the servers, everyone else rents. This concentration excludes researchers, French-speaking startups and the African continent from the AI frontier.

The TOMBOUCTOU innovation. Our architecture is designed RAM-first, optimized for system memory rather than VRAM. Combined with compressed semantic memory and a Mixture of Experts structure, it aims to run frontier-class models on consumer hardware.

The SANKORÉ engine. Inspired by the ijazah, the tradition of certified knowledge transmission inherited from the Sankoré madrasa of Timbuktu, it chains inferences through a persistent semantic memory, for reasoning traceable back to its sources.

✦ Sankoré refers to the great madrasa of Timbuktu (12th–17th centuries), one of the oldest centers of learning in the world. The ijazah was the certification passed from master to student. AfricAIsoft honors this African intellectual heritage by building on its philosophy of rigorous knowledge transmission.

C

Cédric Olivier Somé

Founder · Agentic AI Architect · IT Project Manager

A decade in IT project management, financial analysis and deployment of intelligent systems in regulated environments. That experience ensures a rigorous, compliant sovereign AI deployment that truly serves your organization.

MBA Strategy · Université Laval PMI · Generative AI for PMs PSM I certified FR / EN bilingual
in · LinkedIn profile →
What guides us

Our values

🔍

Transparency

No black box. Source code is delivered, every signal is explainable, every calculation is traceable.

🛡️

Sovereignty

Your data stays on your server. Compliant with Law 25 (Québec) and Law No. 001-2021/AN (Burkina Faso).

🌍

Francophonie

French-first, bilingual, jargon-free. AI turns technical complexity into plain language.

⚖️

Human control

AI proposes, humans decide. No automatic action without explicit approval.

Your data deserves to stay with you.

30 minutes is enough to know whether sovereign AI is right for your organization and leave with a concrete plan. Free, no commitment.