An Unlikely AI Startup Born In Nigerian Hospitals Is Doing What Big Tech Still Can’t

The hospital ward wasn’t quiet. It rarely ever was. Phones buzzed, patients coughed, nurses called out vitals, the agitated and impatient nagged, and through it all, young doctor Tobi Olatunji scribbled furiously, trying to keep up with the flood of patients. Thirty on a light day. Double when things got bad.

It was in that noise—noisy, gritty, chaotic—that Intron was born. Today, that same startup, now rebranded from Intron Health to just Intron, is making voice AI models that reportedly outperform OpenAI, Google, AWS, and Azure when it comes to recognising African accents. Intron stacks up well compared to big names, publicly available benchmarks and datasets reflect.

What started as a solution to medical paperwork has morphed into a robust suite of speech tools, called Sahara, powering voice recognition in hospitals, courtrooms, call centres, and government agencies across the continent.

The premise is simple: Big Tech’s speech tools don’t understand Africa. Intron wants to fix that. But building AI for the hardest accents on Earth didn’t start in a lab. It started in Nigeria’s overstretched clinics, where physicians are lucky to have five minutes with a patient and 30 more filling out forms. Olatunji, now Intron’s CEO, saw that broken system up close and decided to do something about it, through code.

***

It’s easy to romanticise startups. But the earliest versions of Intron didn’t even work well. The first doctors who tried the speech-to-text app during the pandemic took 45 minutes to complete their notes—much slower than writing by hand. Some gave up. Others rolled their eyes. But the problem was real: hospital staff overwhelmed, errors stacking up, and patients at risk.

There’s a particularly haunting story: One Dr. Martins, the only physician at his clinic, missed a biomarker on a routine test. The patient, an elderly woman, had a heart attack a few days later. She survived, but barely. The omission wasn’t due to incompetence. He simply didn’t have time.

It was stories like that—and countless others—that pushed Olatunji and his co-founder, Olakunle Asekun, to go deep on speech recognition. Not just adapting foreign tools, but training new models from scratch.

That led to the creation of AccentMix, Intron’s proprietary algorithm designed to handle one of AI’s thorniest challenges: the wild variability of human speech. So far, Sahara’s models have been trained on over 3.5 million audio clips from 18,000+ speakers across 30+ countries. The result? More than 300 African accents recognised with over 92% accuracy, the company claims.

That isn’t only better than Big Tech on paper but a practical breakthrough. For example, in Nigeria’s Ogun State Judiciary, Sahara has cut court transcription times nearly in half. In Uganda, at C-Care hospitals, patient wait times are down and documentation errors are dropping. Branch International, a notable fintech player, now uses Intron’s conversational bots in its call centres to slash queue times.

And unlike most imported models, Intron’s tools don’t stumble on African names, currencies, or medical jargon. It can transcribe “Ayinla” as easily as “John,” “₦1,250” as smoothly as “twenty dollars,” and understands “troponin” just as well as it does “temperature.”

But perhaps the most interesting part of Intron’s story is how it’s moved from a niche healthtech product to something much bigger: voice infrastructure for the continent.

Earlier this year, Intron launched Sahara-Optimus (its general-purpose voice recognition engine), Sahara-TTS (a pan-African text-to-speech system), and Sahara-Voice-Lock (voice authentication for security use cases).

It’s also training Sahara-Titan, a model that can understand, transcribe, and translate across 20 major African languages including Swahili, Hausa, and Zulu. These efforts have gone from research experiments to products shipping now.

It’s a shift that mirrors how platforms like Google started with search, or Amazon with books. Intron began with hospitals, but the engine it’s building is far more universal. “We built for the hardest environment first,” says Olatunji. “Now, our technology scales effortlessly.”

***

Intron isn’t the only startup working on African voice tech, but not many are doing it at this scale or with this data. And while the team still numbers under 20, the traction is real. Intron now serves 40+ organisations across 8 countries, and its models are deployed in healthcare, justice, finance, and youth health initiatives like Audere’s reproductive chatbot in South Africa.

After a USD 1.6 M pre-seed round in 2024, Intron began expanding both its cloud-native and on-prem deployment capabilities—critical in regions with patchy internet—and growing its engineering and research teams. It also joined NVIDIA’s Inception programme and partnered with the Gates Foundation, Google Research, and Digital Square to benchmark global language models across Africa.

Still, challenges persist. Data collection at scale is expensive. Local hardware constraints remain. And global competition is real. While Intron beats the big names on African voice recognition today, OpenAI, Meta, and Google could close the gap quickly. But this is where Intron’s focus becomes its superpower as Big Tech builds for everyone but Intron is Africa-first.

More than two billion people worldwide are underserved by today’s voice AI. For most of them, English isn’t their first language. For many, the tools built in Silicon Valley don’t even work.

That’s not merely an annoyance but a harbinger of real danger. It means errors in clinical notes. Misunderstood legal testimony. Frustrated customers. Lost time. In places where time is a matter of life and death, that gap can’t be shrugged off.

Intron seems on track to build infrastructure that works for the languages, cadences, and constraints of African life. One dictated sentence at a time.

And while it’s still early days, the company’s trajectory shows what happens when you start with the right problem and build deep. Not to catch up with Big Tech, but to leapfrog it on Africa’s terms.