Problem Solution Family Threads Vision Waitlist
Voice AI for African languages

"My son was born in Belgium.
He doesn't speak Kinyarwanda."

So I built Ijwi AI — the voice layer for Africa.

Diaspora families learn their heritage language. Native speakers contribute their voice. Every conversation trains voice AI for the 2,000+ languages Big Tech left behind.

Africa's languages are invisible to AI.

2,000+
African languages spoken daily
<3%
Have any public voice data at all
0
Big Tech voices that can speak Kinyarwanda

Google can transcribe Kinyarwanda. It still can't speak it.

Two sides. One mission.

Learn

Diaspora families learn their heritage language through voice-first lessons powered by native speaker recordings.

Contribute

Native speakers record their voices to train AI models. Every recording makes the technology smarter.

Every lesson = training data

Three generations. Two continents. One language.

It works like the voice notes your family already sends on WhatsApp — except every message becomes a living lesson, and the language lives on.

Grandma records in Kigali
A voice note, a proverb, a bedtime story — sent the way she already speaks to family.
Her voice becomes the lesson
Ijwi turns her words into prompts, pronunciation guides, and family challenges.
Her grandchild answers from Brussels
The child learns the language. The family keeps the thread. The AI learns real conversation.

The Ijwi Flywheel

Record
Train
API
Revenue
More Data

Each rotation makes the AI smarter.

Building in public.

107+
Prompts
2
Languages
10
Categories
Offline
Mode supported
56
Tests passing
13
Badges to earn

Everyone has read speech.
No one has conversation.

2,384+
Hours of read Kinyarwanda speech exist — public, scraped, already powering Big Tech transcription
Zero
Hours of conversational Kinyarwanda or Kiswahili in any public corpus on Earth

That's what we collect.

Every family thread adds real, emotional, back-and-forth conversation — the data voice agents actually need. It is consented, it compounds daily, and it cannot be scraped.

The voice layer for African daily life.

Africa's informal economy runs on voice, trust, and relationships. The AI economy runs on structured data. Ijwi is the bridge between them.

Daily commerce here doesn't happen in dashboards. It happens in voice notes, WhatsApp chats, and mobile money transfers — in Kinyarwanda, Kiswahili, and 2,000 other languages. The same voice technology that keeps families connected can give that economy a memory.

Speak your business

“I sold 12 crates. Jean paid half. 3 bags left.”

A market seller talks; Ijwi keeps the books. Sales, debts, and stock — recorded in her own language, no forms, no accounting words.

Where people already are

Voice notes. WhatsApp. Mobile money.

No new habits to learn. Ijwi meets commerce in the channels it already lives in — and answers back in the language people actually speak.

Records that open doors

From voice notes to credit-ready.

With explicit consent — and only then — those records can become a business profile a cooperative or lender can trust. The user decides. Always.

$67B
Africa's mobile money market by 2030 — and almost none of it can be operated by voice in the languages its users speak. Family threads today. Economic infrastructure tomorrow.

Your voice stays yours.

Not owned by an institution. Not scraped from the internet. Owned by you, your family, and your community — governed by the Esethu Framework, peer-reviewed African data governance (ACL 2025).

Explicit consent before anything is shared Delete your voice anytime — it's gone Family threads are private by default

Free for African Researchers

Open access to language data for African universities and non-profits advancing local AI research.

Commercial Licensing for Others

Fair licensing for commercial entities. Revenue funds contributor communities and platform growth.

Revenue Reinvested

Profits flow back to the communities who contributed their voices. A virtuous cycle of growth.

Every language deserves a voice.

Stay in the loop. Get updates on launch day.