First season runs November 2026 to May 2027
What it is
Mapinga AI Lab is an African problem-solving laboratory, not a road-technology start-up. It studies one problem at a time, in the place where that problem actually happens, builds the data layer that is missing around it, and turns that layer into decisions people can act on. The lab keeps one method and one subject at a time. The current subject is rural roads; agriculture, water, climate, health, and education will each take their turn only once the current problem has changed a real decision.
The problem
Tanzania's district and feeder road network is the last mile to every farm, dispensary, market, and school — well over 100,000 km of it, most of it unsurfaced, and a large share of it non-engineered earth road that gives way in heavy rain. Road funding is scarce and arrives late, and independent audits show that planned improvements are regularly under-delivered, with spending decisions often following fixed budget ceilings rather than actual road condition. The underlying issue isn't the state of the roads — it's the absence of a continuous record of that state. Condition surveys are periodic and paper-based, and blind to exactly the weeks when rain does the damage.
The approach: invisible AI
The lab's guiding principle is that people should experience solutions that work, not need to understand the technology behind them. A ward officer doesn't need to know an algorithm ranked the risk; they just need a repair list they can trust. That principle sets the order of work: observe the environment and the people first, and only then decide whether — and how — technology helps.
First instrument: Njia
Njia turns scattered observation into a ranked, explainable order of road repair. It draws on trained local observers, vehicles already travelling the routes daily, rainfall and terrain data, and the locations of clinics, schools, and markets. It produces a ranked repair list, the failure pattern and frequency behind each ranking, who and what is cut off by each failure, and the cost of leaving it unrepaired. It is designed for use by ward and district road offices, TARURA engineers, health and education planners, and transporters, insurers, and NGOs who depend on the same routes.
What the lab is honest about
- Volunteer reporting decays over time, so Njia is built around a small, trained, compensated observer corps and vehicles already on the route — wider community input amplifies this, but is never the foundation.
- The underlying data will be noisy, so the system is designed to be useful even at low confidence, and to publish that confidence alongside its answers.
- Evidence alone doesn't move money — Njia's output is shaped to fit directly into the district planning calendar, in a format an authority can act on.
- The data belongs to the community that generated it: public layers stay open, while anything that identifies a person stays within the district.
Proof, not promises
The first season runs with the rains, roughly November 2026 to May 2027, over a defined stretch of the Mapinga road network, and is judged against three concrete proofs:
- 1.A full season's failure events captured and held as a reliable record.
- 2.A ranked repair list actually accepted by a ward or district office into its own planning process.
- 3.At least one repair carried out earlier, or in a different order, because of what Njia showed.
A second experiment: raising problem-finders
Alongside the first season, a small cohort of children in Mapinga will learn to look at their own surroundings and ask what isn't working, why, who it affects, and what a better solution could look like — using Njia's own field data as material, and treating technology as one option among several. The goal isn't children who can operate AI; it's children who can find a problem worth solving.
Who the lab is looking for
- District road and planning officers willing to test a priority list against their own judgement
- Researchers in geospatial and pavement engineering who can show where the model is wrong
- Operators whose vehicles already run these routes daily
- Funders willing to back a laboratory that reports what fails as carefully as what works
Get involved
If you work in road planning, geospatial or pavement research, transport or logistics operations in Tanzania, or fund early-stage African innovation, we'd like to hear from you.
Partner With Us