Soynade Research Charter
Who We Are
Soynade Research is an AI startup dedicated to democratizing technology for West African languages. Our name, "Soynade," is a verb in the Fula language meaning "to see far" or "to look further."
We exist because the sun lights the known world, but the moon reveals the world to come.
While big tech chases "daylight markets", the visible, easy-to-reach users, we bring moonlight to the digital divide. We illuminate the landscapes others perceive as shadows. To "Soynade" is to possess a dual power:
The Sharp Eye (Precision): Like a guide spotting a path in the dark, we have the visual acuity to see the nuances of oral African languages and the specific offline needs of our communities.
The Long Look (Foresight): We look beyond immediate obstacles. We see past the "barriers" to a glowing horizon where African technology is independent, open, and accessible.
At Soynade, we see differently.
Our Vision
To see further than the horizon. We envision a digital future where every African language is a first-class citizen of the internet, supported by open, offline-capable, and locally grounded AI technologies.
Our Mission
We build the infrastructure for African digital independence. Our work focuses on developing open-source Large Language Models (LLMs), Speech Language Models, speech recognition systems, Vision Language Models, and translation tools tailored for West African languages.
We prioritize:
Orality: Because our cultures speak before they write.
Offline Accessibility: Because connectivity should not be a barrier to intelligence.
Community-Driven Development: Because technology must serve the people, not the other way around.
Community Principles
Our open-source community is guided by three core tenets:
The Moon Shines for All
No one owns the moon; it shines for the shepherd and the king alike.
Our technology is built to be accessible to everyone, from the university researcher in Dakar to the farmer in rural Mali, from the developer in Lagos to the teacher in Ouagadougou.
See Farther Together
We value contributions that extend the reach of our tools, whether through code, data, or feedback. We believe that when we share our "vision," the whole community sees further.
Cultural Dignity
We treat African languages and data with dignity. We reject "data mining" mindsets. We ensure accuracy, context, and cultural relevance in all our datasets and model outputs.
Licensing Strategy
Our licensing reflects our commitment to sustainable open source. We believe that those who benefit from our research should contribute back to the ecosystem.
Code, Tools & Pipelines: AGPL-3.0 (Affero General Public License v3.0)
We treat our core R&D-including data processing pipelines, training frameworks, and infrastructure tools-as shared community resources. Released under the AGPL-3.0 license, they require that any modifications or improvements, including when used to provide network services or APIs or internal production modifications, be publicly released under the same terms. This ensures that commercial and cloud-based deployments built on our work contribute back to the community.
New Models: AGPL-3.0
All LLMs, speech models, and vision-language models are released under the AGPL-3.0 license. This includes model weights, architectures, and training code. Any modified, fine-tuned, or derivative versions must also be distributed under the same license, including when deployed as network services. This prevents the privatization of common goods by requiring that improvements built on our work remain publicly available.
Datasets & Benchmarks: CC-BY-SA 4.0
Our high-quality training corpora and "gold standard" evaluation sets are community assets. We do not impose any terms except those required by the CC-BY-SA 4.0 license. But we encourage that any enriched, corrected, or augmented versions of these datasets be shared back to the community.
Legacy Models: Apache 2.0 (Oolel v1 and earlier releases remain under their original permissive license).
Why AGPL-3.0?
Community Sustainability: Companies benefiting from our work must either contribute back or negotiate commercial licenses, creating a sustainable ecosystem.
African Digital Independence: Prevents large tech companies from building proprietary services on our research without giving back to African communities.
Dual Licensing Opportunity: Enables us to offer commercial licenses to organizations that cannot comply with AGPL terms, supporting our mission financially.
Commercial Licensing
Organizations that wish to use Soynade technology in proprietary products or services without releasing their modifications can contact us about commercial licensing arrangements. Revenue from commercial licenses directly funds further development of our open-source tools.
Trademarks
While our code is open, our brand identity is protected to ensure trust:
Soynade™: The project name, logos, and brand identity.
Oolel™: The Wolof LLM family name.
Firtanam™: The Fula translation tool family name.
Contact
For all inquiries, issue reports, or partnerships, reach out to: contact@soynade.ai