A Fresh New Face for Samasource

Last year, Samasource celebrated its tenth birthday, and today, we celebrate a new brand, centered on our belief that training data is the soul of AI, and that the highest-quality, most secure training data is produced by a social impact business model centered on measurable poverty reduction.

I started Samasource in 2008 when I was 25 filled with a single big dream: show the world that young people from low-income backgrounds in East Africa could not only work in tech, but transform it and move out of poverty in the process. I’d been passionate about working on social justice in sub-Saharan Africa since I was 17, when I left high school early to work in Ghana as an English teacher and became frustrated by the traditional aid approach. Instead of giving people handouts, which they didn’t want, and promoting the “white savior” mentality that had insidiously worked its way into the development narrative (just calling these countries “Third World” implied that it was charitable to spend time in them), why couldn’t we give work instead? It seemed to me that the best way to promote social justice in the region was to create dignified, living-wage jobs for the poorest people.

So off I went. I started Samasource as a non-profit organization because no one believed it could ever be a business. I’d modeled our organization after Muhammad Yunus’s Grameen Bank — I wanted to hire only people from low-income backgrounds (with the further condition that at least 50% of them had to be women), to train them in tech skills, hire them to work on contracts I’d secured, and pay them living wages so they’d move out of poverty on their own terms.

This was all too much — both for Sand Hill Road, and I discovered, for large foundations. One person at a foundation considering us for a grant told me, authoritatively, “Poor women in Africa need mosquito nets before they can learn to use a computer!” (This, as the continent was surpassing the US in mobile phone usage.)

So we had to prove them wrong. I slowly won contracts, hired up, and we grew through an earned-revenue model. We started building training data for AI teams in Silicon Valley in the very early days. Our first projects, annotating points on images for Microsoft in 2012, were delivered at over 99% accuracy by our team of over 300 in Kenya. We won grant support and over time I reorganized the company as a hybrid with for-profit subsidiaries operating parts of our business under a nonprofit owner, so that we could take in investment capital to scale further.

It was an unusual model (a Silicon Valley company that’s owned by a nonprofit, and founded to reduce poverty?), but the approach worked. We’d found our niche in training data and by the end of last year, we’d served 25% of the Fortune 50 while moving over 10,000 agents out of poverty, increasing their incomes by over 500% with our living wage model. Last year alone, our workforce annotated over 1B points on images, video, and LiDAR output — fueling AI for autonomous driving, smart devices, and even conservation (wherein our agents train algorithms to recognize rare species or conduct animal censuses from video feeds).

We now employ over 2,000 people in East Africa and have paved the way for a digital work boom in the region. With new management training, financial literacy, and powerful digital skills, many of our agents go on to found their own companies. And our messaging is finally catching up, with a new marketing team and technology leadership powering a new vision for the next decade of growth. I’m proud today to announce a new brand for Samasource, focused on training data as the soul of AI, and continuing our commitment to scalable social enterprise.