Who's building this, and why it's built this way.
To build Africa's fully-equipped AI workstation — the single place every African decision-maker, from farmer to founder, gets intelligence built from Africa's own ground truth.
Not a chatbot wrapped in African prompts, and not a slogan — an infrastructure layer, built one verified word and one earned specialist at a time.
Mark Gakuya
Bootstrapped AfriFoundry Limited solo from Mombasa, Kenya, building the company alongside a Biomedical Engineering degree. AfriFoundry exists because Africa's languages and markets don't yet have AI infrastructure built from their own ground truth — Afri3B is the attempt to build it honestly, one verified word at a time.
Get in touch▾▴
Where this actually started
AfriFoundry's first product was a general-purpose assistant wrapping existing AI models with African framing. It worked, but it wasn't building anything of its own — every answer still came from somewhere else. That product was retired. Afri3B is the deliberate rebuild: infrastructure trained from real African ground truth, not adapted onto it afterward.
The team, growing
AfriFoundry has been built solo from day one. A team is now being brought on to build at the scale this vision actually needs.
Meet the team, and who we're looking for →Never fabricate
If the data doesn't exist yet, Afri3B says so — it doesn't invent an answer to sound complete.
African ground truth first
Kenyan markets, real languages, real usage — the foundation, not a translation layer over someone else's model.
Earned, not claimed
A capability is only real once it's actually built and tested — not announced ahead of it.
Built in Africa, for Africa
Decisions get made from Mombasa and Nairobi, for the markets AfriFoundry actually knows.
Built one verified step at a time.
June 2025 — AfriFoundry founded
Bootstrapped solo from Mombasa, alongside a Biomedical Engineering degree.
May 2026 — First live partner integration signed
A production API integration agreement with a Kenyan multi-marketplace and logistics platform — the proof point behind the Partners page.
Ongoing — Language infrastructure in active build
Self-building dictionary and hybrid routing live; more than ten African languages in active ingestion and training.
Straight answers, since this matters to us.
Is Afri3B live right now?+
Afri3B is live as a hybrid system — general-purpose reasoning wrapped in African context today, with Afri3B's own trained models taking over language by language as each one earns it. Nothing is claimed as "fully trained" before it actually is.
What languages does it support?+
English and Swahili are handled natively today, including code-switched queries. More than ten additional African languages are in active ingestion and training.
Where does the data come from?+
Real ingested sources only — manual collection, partner data, and verified corpora — never scraped and fabricated to look complete. If a word or pattern isn't verified, it's flagged, not guessed.
How is this different from wrapping GPT or Gemini with African prompts?+
That's the starting point, not the destination — every domain is built toward its own trained weights from real African data, with the general-purpose layer only ever a honest, temporary scaffold.
Who's behind AfriFoundry?+
Mark Gakuya, founder and CEO, building from Mombasa and Nairobi, Kenya.