A physical AI presence for Africa

Amai.Mother.

Shona · noun · the one you go to

AMAi is a live African learning platform. Teachers, learners and examiners use it today. We are giving it a body: a solar unit you walk up to and talk to, in your own language, with no smartphone, no data bundle and no account.


The name is the thesis

In Shona you do not call your mother by her name.

You call her Amai. It is a term of address, not a label: it marks respect, and it marks who you are allowed to bring a problem to. Mashiri's work on Shona address terms describes exactly this, that kinship terms carry social obligation, not just identity.

We named the platform AMAi before we understood how much work the name would do. A mother does not ask which school you attend, whether your fees are paid, or whether you own a phone. That is the product specification.

Pedzisai Mashiri, "Terms of Address in Shona: A Sociolinguistic Approach," Zambezia, 1999.

A NAME one person points at an individual AMAI an obligation anyone may invoke it and be answered The unit answers to the term, not to a customer record.
Why the name matters commercially: a kinship term implies a duty to answer. That is the behaviour we are engineering for.

Why a body, and not another app

The learners who need this most cannot reach an app.

Every number below is the reason the phone-shaped version of AMAi stops at the edge of town.

14%
of rural Zimbabwean households have electricity
Each dot is one household in a hundred. The unit runs on the sun so it does not need the other 86.
Has a phone 96% Has the internet 38% 58 points of people reachable by voice, not by app
96% / 38%
mobile penetration against internet penetration
The gap is the market. A voice unit needs neither a smartphone nor a bundle.
16 OFFICIAL LANGUAGES 1 partial · 0 production-ready
1 / 0
usable Shona models, and Ndebele models, that exist today
We are not licensing this from anyone. Speech in these languages is a thing we have to build.
SOLAR CANOPY speaks at chest height no account no handset no bundle
Concept illustration. Industrial design is not finalised; the constraints shown, solar power, chest-height voice, a screen that is optional, are.

Amai, with a body

You walk up. You talk. She answers in Shona.

The unit stands where people already gather: a clinic yard, a school gate, a growth-point market. It runs on the sun, caches its models locally, and syncs when a signal appears. It does not need to know who you are before it helps you.

Everything it says is produced by the same engine that already marks exams and runs seminars for real Zimbabwean teachers today. The body is new. The brain is in production.

Sixty seconds at the unit

What actually happens when someone walks up.

00:00

She is just standing there

No screen to unlock, no queue, no one to ask permission from. The ring is dim until someone is close.

"Amai, ndibatsirei..."
00:04

He asks in Shona

Speech recognition runs on the unit itself. No bundle is consumed, and nothing has to be typed.

00:11

She answers out loud

Voice first, because literacy cannot be assumed. The screen is a courtesy for anyone who prefers to read.

syncs when a signal appears
later

A teacher sees it

The same platform that marks exams picks up what the unit heard, so the interaction is not thrown away.

The engine is not a plan

AMAi is already running in production.

This is the part most hardware pitches do not have. Teachers, learners and examiners use these surfaces today. The unit is a new front door onto a system that already works.

Live

Exam practice and marking

Learners upload papers; AI and human markers both score them, and results come back with a candidate number.

Live

Seminars

Facilitators run live sessions with enrolment, capacity and attendance built in.

Live

Homework and revision

Teachers set work, learners submit, and the platform tracks marking through to a grade.

In build

Shona and Ndebele speech

An evaluation lab is running today. Production-quality recognition in both languages is what this raise funds.

Crowdfunding the first units

What your money actually buys.

We are funding a specific, boring, checkable list: speech data collection, one hardware revision, and the first units in the ground.

$25

An hour of recorded speech

Native Shona and Ndebele speakers, paid properly, recorded in the conditions the unit will actually face.

$250

A day of the build

Enclosure, power budget, thermal work. The unmarketable engineering that decides whether it survives a year outdoors.

$5,000

One unit, sited and named

A unit in a community you can name, with public reporting on what it was asked and whether it helped.

Before you give us anything

What happens to the box if we go bust?

Things we would rather you heard from us

  • No unit exists yet. The illustration above is a concept. The software behind it is real and in production; the hardware is not built.
  • Production-quality Shona speech does not exist yet, from us or anyone. That is the single biggest technical risk in this pitch, and the main thing the money is for.
  • If we fail, the units keep working offline until their models go stale, and we publish the speech data and firmware so somebody else can carry it. That commitment is written into the raise, not added afterwards.
  • This is a crowdfund, not an equity offering. You are backing a build, not buying a security, and you should not treat it as an investment.
  • We are a small team in Zimbabwe and the United States. Timelines will move. We will say so when they do.