Lesan

Some languages around the world have tens of millions of speakers, but AI-powered translation tools like Google Translate did not work well for them. Lesan set out to solve that: if we could build excellent and large datasets of parallel corpora, we could power strong machine translation models and open up the web's content to millions of people.

I went through Entrepreneur First and met my co-founder there, a very talented machine learning researcher. We lived in Ethiopia, built an operational team on the ground, and produced SOTA machine translation models for Amharic and Tigrinya. We went through Meta's Connectivity Accelerator and to date we have completed millions of translations for thousands of users worldwide.