ProjectVoice & Language DataIn development

Local Language AI for Education

Many children in Uganda learn at home in one language and at school in another. Mathematics and Science terms then arrive in English. Neuravox builds AI for learning in local languages, so that a child can ask a question in Luganda, hear the answer in Luganda, and connect it to the English term they will meet in class and in examinations.

What we are building

Our first programme covers Primary One to Primary Five. It supports early numeracy and science in the first three years, and Mathematics and Integrated Science in Primary Four and Primary Five.

The system answers only from approved curriculum content, not from the open internet. It uses a reviewed Luganda terminology layer for mathematical and scientific concepts. It draws on reviewed Luganda proverbs, sayings and everyday examples to connect lessons to the lives of children. It accepts spoken questions and gives spoken answers in Luganda. The teacher stays in control: teachers accept, edit or reject every activity the AI writes before they use it in class.

How we work

We adapt existing language and speech models. We do not train new foundation models. We put our effort into the parts that decide quality in a classroom: grounding in the curriculum, review of terminology, cultural review, speech testing with teachers and pupils, and data protection for children. Every source in the system has a recorded owner and permission. We do not keep the voices of pupils after processing.

Status

In development with an education partner in Uganda. School validation is planned for late 2026.

Next

We plan to extend this work to more subjects, more classes and more languages across Africa. We welcome partners in education, language research and funding.