We are building data and software infrastructure to ensure artificial intelligence serves local communities reliably.
AI models trained on Western datasets frequently break down when faced with the unique physical, cultural, and environmental realities of diverse regions. Here is how we bridge that gap.
Western autonomous driving models are trained on structured highways and orderly lanes of Silicon Valley. If placed in Mbare (Harare's busiest, most chaotic market suburb), they encounter street vendors, pushcarts (matsimba), informal minibus taxis, jaywalkers, and unstructured traffic flows that the perception system has never seen. BTech curates high-fidelity visual datasets of these local scenes to train reliable computer vision models.
Many large language models translate Shona words literally but miss the tone, accent, and cultural contexts. The resulting speech interfaces sound artificial, use incorrect accents, or misinterpret regional metaphors. This is because they are trained on scraped web data without native-speaker QA. BTech curates conversational voice and text datasets to build natural Shona language interfaces.
We do not have any open roles at the moment, but we are always looking for exceptional software developers, machine learning engineers, and data annotators. Send your CV and portfolio to us for future consideration.