Smart agriculture
Season forecasting, crop disease detection, irrigation optimization. Our models adapt to tropical climates and local crops.
AI for Africa · Local solutions
Smart agriculture, assisted medical diagnosis, personalized education, inclusive fintech. We build AI solutions tailored to African realities, in French and local languages.
Find your use case
A single photo of the leaf is enough; the analysis runs on the phone.
Models tuned to tropical climates and local crops.
Sensors and AI to decide what to plant, when, and how much to water.
An editorial selection of cases genuinely deployable across the continent — not a promise of measured results.
Where AI makes the difference
Concrete solutions, tailored to local infrastructure and needs.
Season forecasting, crop disease detection, irrigation optimization. Our models adapt to tropical climates and local crops.
Assisted diagnosis, patient monitoring, telemedicine. AI helps offset the shortage of specialist doctors in rural areas.
AI tutors in French and local languages, automatic grading, adaptive learning paths. Quality education accessible to all, even without a permanent connection.
Alternative credit scoring, fraud detection, conversational agents in local languages. AI opens access to financial services for the unbanked.
Speech recognition and translation in Wolof, Bambara, Hausa, Swahili. Our models understand linguistic and cultural specifics.
Electricity demand forecasting, predictive grid maintenance, solar mini-grid optimization. AI makes infrastructure more resilient.
How we work
We work with local experts to understand real needs, infrastructure constraints and cultural specifics.
Lightweight models that run on modest devices, tolerate intermittent connectivity, in French and local languages.
We train local teams and provide ongoing support. Our solutions are built to be maintained by African talent.
Our principles
Intellectual property and execution stay African. Our tools are designed to be maintained locally.
Lightweight models that run on modest devices and tolerate an intermittent connection.
French isn't enough: we integrate the languages people actually speak, written and spoken.
Every project is built with field experts, NGOs, universities and the beneficiaries.
What you want to know
Yes, our models are designed to work offline or with reduced bandwidth. Data is processed locally on the device, and synchronization happens when a connection is available.
We work with anthropologists and local experts. Our models are trained on African data and tested by local users before deployment.
We offer models adapted to local budgets, starting from €1,000 for a basic solution. Funding and partnerships are available for high social-impact projects.
We run a free training program for African developers. Our solutions are open-source and documented in French and English. We hire locally.
Feasibility simulator
4 questions to get a personalized recommendation: deployment type, estimated complexity, key challenges and a concrete first step.
Model comparator
About twenty lightweight models genuinely usable on modest devices (< 2 GB RAM), filterable by task, RAM, language and sector. Natural complement to the feasibility simulator: it tells you which deployment type, this tool answers 'which specific model to use'.
20 models match
| Model | Task | Size | Min RAM | Supported languages | Typical use case | |
|---|---|---|---|---|---|---|
| Whisper tiny | 🎤 Voice | 39 MB | 512 MB | 99 langues / 99 languages | Offline speech transcription: meetings, field notes, dictation | View → |
| Whisper base | 🎤 Voice | 74 MB | 512 MB | 99 langues / 99 languages | Higher accuracy, multi-speaker, varied accents | View → |
| MMS-TTS (Meta) | 🎤 Voice | 100 MB | 512 MB | Hausa, Yoruba, Swahili, Amharic… | Text-to-speech in African languages without internet | View → |
| AfriBERTa | 📝 Text | 110 MB | 512 MB | Yoruba, Hausa, Amharic, Shona, Igbo… | Text classification and sentiment analysis in African languages | View → |
| MasakhaNER | 📝 Text | 110 MB | 512 MB | Amharic, Hausa, Igbo, Yoruba, Swahili… | Named entity extraction: places, people, organizations | View → |
| YOLOv5 nano | 👁 Vision | 4 MB | 512 MB | — | Real-time object detection on modest devices | View → |
| MobileNetV3 | 👁 Vision | 5 MB | 512 MB | — | Lightweight image classification, visual diagnosis | View → |
| PlantDisease CNN | 👁 Vision | 50 MB | 512 MB | — | Crop disease diagnosis from photos, works offline | View → |
| EfficientDet-Lite0 | 👁 Vision | 4 MB | 512 MB | — | Ultra-light detection, Edge AI on entry-level devices | View → |
| ResNet-50 | 👁 Vision | 98 MB | 512 MB | — | General-purpose visual classification: crops, pathologies, infrastructure | View → |
| Qwen2-0.5B | 📝 Text | 390 MB | 512 MB | ZH + EN | Text generation, lightweight QA, mobile reasoning | View → |
| SunbirdAI ASR | 🎤 Voice | 200 MB | 512 MB | Luganda, Swahili, Kinyarwanda | Speech recognition for East Africa | View → |
| TinyLlama-1.1B | 📝 Text | 637 MB | 1 GB | EN (partiel multilangue) | Conversational agent, summarization, instruction following | View → |
| SmolLM2-360M | 📝 Text | 720 MB | 1 GB | EN | Text summarization, classification, lightweight on-device assistant | View → |
| Phi-1.5 (Q4_K_M) | 📝 Text | 832 MB | 1 GB | EN | Advanced reasoning, coding, on-device QA | View → |
| mGPT | 📝 Text | 800 MB | 1 GB | 61 langues (FR, AR, langues slaves…) | Multilingual text generation including French | View → |
| AfroXLMR-base | 📝 Text | 900 MB | 1.5 GB | 17 langues africaines | African NLP: classification, QA, information extraction | View → |
| BLOOM-560M | 📝 Text | 1100 MB | 1.5 GB | FR + EN + 46 langues | Generation in French and African languages (Wolof, Bambara…) | View → |
| NLLB-200-distilled-600M | 📝 Text | 1200 MB | 1.5 GB | 200 langues dont 50+ africaines | Translation to/from Fulani, Bambara, Lingala, Dioula… | View → |
| Gemma-2B (Q4_K_M) | 📝 Text | 1340 MB | 2 GB | EN (multilingue partiel) | Advanced comprehension, long summarization, on-device AI agents | View → |
Offline speech transcription: meetings, field notes, dictation
99 langues / 99 languages
View →Text-to-speech in African languages without internet
Hausa, Yoruba, Swahili, Amharic…
View →Text classification and sentiment analysis in African languages
Yoruba, Hausa, Amharic, Shona, Igbo…
View →Named entity extraction: places, people, organizations
Amharic, Hausa, Igbo, Yoruba, Swahili…
View →Conversational agent, summarization, instruction following
EN (partiel multilangue)
View →African NLP: classification, QA, information extraction
17 langues africaines
View →Generation in French and African languages (Wolof, Bambara…)
FR + EN + 46 langues
View →Translation to/from Fulani, Bambara, Lingala, Dioula…
200 langues dont 50+ africaines
View →Advanced comprehension, long summarization, on-device AI agents
EN (multilingue partiel)
View →Sizes are for quantized variants (GGUF Q4 / int8) — always verify on the official model page. All work offline.
Contact
Have a project? Want to learn more? Get in touch with our team.
Response within 48h · Free quote