AI intellectual property and research infrastructure — building South Africa's first official Large Language Model, trained on all eleven official languages.
Most models meet South Africa in translation. AURA meets it in the mother tongue. We are building the country's first official Large Language Model — pretrained and aligned across all eleven official languages, so reasoning happens natively rather than through an English bottleneck.
AURA is the research spine of the group. It powers SCHOOL's AI tutor, vets Cryptic SPHERE's trading signals, and keeps the intellectual property — the data, the weights, the value — owned in South Africa, for South Africa.
South Africa's first official LLM — pretrained and aligned on all 11 languages, not translated from English.
Owned compute and data pipelines — the substrate every Ditshego venture builds intelligence on.
AI intellectual property held in South Africa, for South Africa — data, weights, and value kept home.
The tutor behind the flagship — AURA reasons in the learner's mother tongue.
Speech models for low-bandwidth, multilingual, real-world classrooms and call centres.
Published benchmarks and datasets to seed an African AI research community.
"A nation that does not own its intelligence rents its future. We intend to own ours."
