A student project focused on collecting speech data in Zimbabwean languages was among the standouts at the 2026 AI for Impact Challenge. Inzwi, developed by Midlands State University students Nigel Basarokwe, Tadiwa Njimaima and Thabelo Dumani, is described by the university as a WhatsApp- and web-based platform that crowdsources paired text and audio, beginning with Shona and Ndebele.

The challenge, hosted by the Postal and Telecommunications Regulatory Authority of Zimbabwe (POTRAZ), drew 738 applications, with 60 innovators selected for a residential bootcamp in Nyanga. MSU reported that Inzwi received first place in its development domain and a prize package valued at US$21,000, made up of AI tokens and a cash award. The university says the project began as Basarokwe’s final-year research into the shortage of high-quality speech datasets for local languages. (Midlands State University Research and Innovation Division; NUST report on the challenge)

Why speech data matters

Voice assistants and speech-recognition tools need recordings that reflect how people actually speak, along with accurate transcripts. For languages with less material available for training and evaluation, that data can be difficult to assemble. A collection tool built around local participation could help researchers and developers create resources for speech technology that better represents Zimbabwean speakers.

That possibility depends on careful data stewardship. Participants should understand how recordings are collected, stored, licensed and used; consent and privacy protections matter as much as the technical pipeline. A growing dataset is useful only if the people contributing to it are treated fairly and its quality is checked.

From competition result to sustained resource

The challenge gave Inzwi visibility and access to compute support, but an award does not yet establish that the platform is operating at scale or that its data has improved deployed speech systems. MSU’s account identifies the project’s direction; the next milestones are a functioning collection process, transparent governance and datasets that other researchers can actually use.

Inzwi points to a concrete role for local AI work: building the data foundations that make later tools possible. The long-term test is whether the project can move from a student prototype into a trusted, maintained resource for Zimbabwean language technology.