The global ed-tech conversation tends to assume that students have internet access. AI tutors, learning management systems, video platforms, and adaptive testing all depend on connectivity. But the International Telecommunication Union estimates that 2.6 billion people remain offline worldwide. For the hundreds of millions of school-age children among them, the digital education revolution is happening on a different planet.
Learning Equality, a nonprofit based in San Diego, built Kolibri to bridge this gap. Kolibri is an open-source educational platform designed from the ground up to work without the internet. It runs on a local server, a low-cost device like a Raspberry Pi, a repurposed laptop, or even a single tablet, creating a local network that students can connect to with any device. The content library includes materials from Khan Academy, CK-12, PhET simulations, and dozens of other providers, all curated and formatted for offline use.
200 Countries and Counting
Kolibri has been deployed in more than 200 countries and territories. Its users include students in rural sub-Saharan Africa, refugee learners in camps across East Africa and the Middle East, and indigenous communities in South America and Southeast Asia. UNHCR, UNICEF, and World Vision are among the organizations that have incorporated Kolibri into their education programs.
The platform's reach is particularly significant in fragile and conflict-affected contexts. In refugee camps, where infrastructure is temporary and internet connectivity is either unavailable or prohibitively expensive, Kolibri provides a way to continue education. A single server loaded with content can serve an entire school. Content updates can be delivered on USB drives during periodic connectivity windows, then distributed across the local network.
Adaptive Learning Without the Cloud
What makes Kolibri more than just an offline content library is its built-in learning management and recommendation system. Teachers can assign lessons, track student progress, and view analytics on class performance. The system uses algorithms to recommend content based on a student's demonstrated mastery level, creating a personalized learning path even without cloud-based AI.
Learning Equality has been developing lightweight AI features that run entirely on local hardware. These include automated assessment, content recommendations based on learning patterns, and early warning systems that flag students who are falling behind. The models are small enough to run on a Raspberry Pi, designed for the constraints of offline environments where computing power is measured in fractions of what a modern smartphone provides.
Evidence of Impact
Rigorous impact evaluations have been conducted in several deployment contexts. In Cameroon, a study found that students using Kolibri achieved a 14% improvement in math scores compared to a control group. Across multiple sites, students showed a 36% improvement in creative and critical thinking skills as measured by standardized assessments.
In India, a deployment in partnership with Nalanda Project reached thousands of students in rural schools across several states. The evaluation found that Kolibri was most effective when combined with teacher training, reinforcing the principle that technology works best as a complement to human instruction, not a replacement for it.
Among refugee populations, the outcomes are harder to measure quantitatively but no less important. For displaced children who may have been out of school for years, access to structured educational content in their language can be the difference between continued learning and a generation lost.
The Content Challenge
One of Kolibri's most important innovations is its content pipeline. The Kolibri Content Library is a curated collection of open educational resources, organized by subject, grade level, and language. The platform supports over 30 languages, with content ranging from primary school math to secondary school science.
Content curation for offline use is more complex than it sounds. Videos must be compressed. Interactive simulations must be adapted to run without web services. Metadata must be standardized so that the recommendation system can function across content from different providers. Learning Equality has built an entire tool chain, called Kolibri Studio, that allows content creators and local education organizations to package and distribute materials for the platform.
Closing the Last Mile
The ambition behind Kolibri is simple to state and enormously difficult to execute: ensure that the accident of being born without internet access does not determine a child's educational future. In a world where AI-powered education tools are increasingly powerful but increasingly dependent on connectivity, Kolibri represents a counter-current. It insists that the most marginalized learners should not be the last to benefit from educational technology, but the first.
Two hundred countries. Refugee camps and rural schools. A Raspberry Pi running AI recommendations for a classroom full of students who have never seen the internet. This is what educational equity looks like at the hardest edge of the problem.
Sources: Learning Equality official deployment data and impact reports; UNHCR education technology programs; ITU "Facts and Figures" connectivity data (2024); Kolibri impact evaluation in Cameroon (2023); Nalanda Project India implementation reports; Kolibri Studio documentation.
