Before this moment, speech-to-text technology—largely developed in the laboratories of the West—remained deaf to the specific inflections of Nigerian English. Radiologists often found that standard software would stumble over their vowels or fail to recognize terms for endemic conditions like Lassa fever. Dr. Olatunji, a physician who had previously managed machine learning for Amazon Web Services, recognized that this was not a failure of the users, but a failure of the data. He set out to build a digital ear that understood the continent’s voices.
The challenge was one of volume and variety. Olatunji and his team at Intron Health systematically built a proprietary dataset by recording healthcare workers as they read medical terms aloud. They sought out the specific rhythms of speech that commercial tools ignored, ensuring the vocabulary included locally available medications and regional terminology that are frequently absent from global dictation libraries.
To make the system functional within the reality of a busy hospital, the developers even accounted for the ambient environment of the wards. The software was engineered to filter out the constant, rhythmic whir of the ceiling fans, allowing the doctor’s voice to remain clear despite the heat and the noise of a public institution. It was a technical solution born from an intimate understanding of the setting in which it would operate.
The result is a liberation of time. In a country where only 35,000 doctors serve more than 200 million people, the burden of documentation has long been a barrier to care. By shifting from handwritten notes to instant transcription, the hospital has effectively reclaimed hours of clinical focus. For the patient waiting in the hallway, the innovation means the difference between a diagnosis that takes two days and one that arrives before they have even left the building.