If you’re not very familiar with the African AI ecosystem, you might expect to venture in and just find people using AI. What you’ll be certain to find, alongside that, are people building it. Not frontier building, but the more immediate kind of identifying problems that global tools were not designed to solve, and solving them.
When YPIT spent the better part of a year mapping Nigeria’s emerging AI ecosystem, speaking to engineers, researchers, founders, and practitioners, that’s exactly what they found – A cluster of people doing good, hard work in the field of AI. This cluster is small enough that many of them know each other, and their work, specific enough to be worth examining.
Taking a deeper look across the companies that have emerged from this ecosystem, a pattern becomes visible. Companies like Tonative are working at the data layer, working hard at curating African language datasets. Without this raw linguistic material, any model trained for African contexts is operating on low-resource data. It’s unglamorous work, rarely discussed, but it’s foundational.
On top of it, companies like Spitch and Intron are building the model layer. Spitch is building speech recognition and text-to-speech for Yoruba, Hausa, Igbo, and Nigerian-accented English, amongst others. Intron Health is doing similar work in healthcare with Sahara: a speech recognition model covering 20-plus African languages and accents, with offline functionality and live deployments in Nigerian and Kenyan courts and hospitals.
TechCabal has covered both startups extensively. Companies like Oyster make the same argument in skincare. They train their AI on African and darker-skin datasets, correcting a specific failure: global skincare models perform significantly worse for most African users because the training data did not include the skin it was meant to serve.
At the infrastructure level, companies like Cencori and Yamify, are building compute access and deployment infrastructure at scale across Nigeria and Congo. TechCabal Insight’s own State of Tech in Africa 2025 found that less than 5% of African AI talent has access to the GPU power needed to build seriously. Yamify abstracts that gap behind a simple interface, deploying in African data centres. Okalobe’s framing: “You are a renter of intelligence. You are not a landlord. We want Africans to be landlords.”
On the surface, you might think these are isolated experiments. But they’re not. They are components of a stack: data, models, infrastructure, being built in parallel by people who understand what the gap looks like for Africans, because they have had to play in this space.
The implication hiding in plain sight
I believe the primary reason this cluster of companies exists is necessity, before ambition.
The AI tools built at scale were built for the contexts their builders understood. If you take it as criticism, it’s not. It’s just how tools work. A language model trained without African language data will not understand African language. A skincare AI trained without African skin data will not recognise African skin conditions reliably. A speech recognition tool built for hospital environments will not perform well in a Lagos health clinic. There should be no surprises there.
What this means though, is that if the next decade of applied AI is built without African participation, the tools that govern African healthcare, finance, language, and infrastructure will have been built by people who have never needed to solve those problems. Once again, it is how tools work. I suspect that each of the builders above arrived at this same conclusion through experience, not ideology. At some point, they each reached for an existing tool and found that it did not fit our African use cases. And then, they went on to build the solutions.
The pipeline problem
In mapping Nigeria’s AI ecosystem, we found that when it comes to AI talent i.e., people who can credibly tackle problems like the above, the cluster is real. That cluster is also very small.
Ayomide Odumakinde’s path from OAU to Cohere is instructive because of how exceptional it was, not in spite of it. He spent two and a half years after graduation teaching himself mathematics and machine learning from textbooks. He applied to Cohere’s scholars programme and was one of six selected from three thousand global applicants. All of that effort created the opportunity for an exceptional young man to enter into a community of people working on similar problems.
But a community is not a pipeline. AI Saturday Lagos (now TRI AI Saturdays), is a more formal community with more than 7000 members. They run a sixteen-week free AI/ML class, opening up AI education for Africans one Saturday at a time. TRI AI Saturdays exists because its founders saw the need for a pipeline as far back as 2018 when they, themselves recognised the potential of AI to positively impact their society. The community has been covered by TechCabal. This year, they are set to run their 10th AI/ML cohort in collaboration with Google DeepMind and have received more than 2800 applications across more than 80 countries across the world. This is another example of individual initiative, turned community initiative, circumventing institutional design and is a reflection of how talent is nurtured here. An informal community can sustain a cluster but it cannot produce the scale of participation needed to shift Africa’s position in applied AI in any structural way.
For that sort of challenge, we need something more deliberate. We need to stop relying on the accidents of geography, timing, and access that produced these builders above as our primary path to innovation. What we need is a flywheel in which talent becomes visible, visibility creates connection, and connection accelerates problem-solving.
What an intentional response looks like
The Artificial Future is YPIT’s attempt at an answer. Running across April, May and June in Lagos, it brings together a free Webinar and Workshop Series, a two-weekend AI hackathon with 250 builders across 50 teams, and a conference at The Civic Centre on June 13th, designed to function as the kind of structured entry point the builders above had to find for themselves.
The organising philosophy is counterintuitive: high ambition and radical inclusion don’t need to be in conflict. The hackathon covers four tracks, Economic Access, Local Language and Culture, Education, and Healthcare, each a direct reflection of the applied problems the ecosystem above is already working on. A first-year student and a senior ML engineer can both belong in the room. On Demo Day, they will be in the same room as the investors and founders at The Civic Centre.
The builders we’re looking for are already here. We just need to build the infrastructure around them, and we need to do it now.
—
The Artificial Future runs May to June 2026 in Lagos and online. The AI Hackathon takes place May 30 to 31 and June 5 to 6. The AF Conference is June 13 at The Civic Centre, Lagos. Details at www.theartificialfuture.com.
