Why education software fails in Africa
Most LMS platforms treat internet connectivity as a given. Moodle, Canvas, Google Classroom — they all assume that the student's device is online when they need to access content, submit an assignment, or take a quiz. In a semi-urban Nigerian school, a Southeast Asian rural classroom, or an NGO-run learning centre, that assumption fails multiple times per week.
EduCore AI was designed from first principles around intermittent connectivity. The offline experience is not a degraded fallback — it is the primary mode.
Offline-first architecture
All critical learner interactions — viewing course content, completing exercises, submitting assignments, taking quizzes — work fully offline. The sync layer uses WatermelonDB with a custom conflict resolution matrix:
- Course content: Teacher writes always win on sync. Learner reads are append-only.
- Assessment attempts: Learner writes are sacred — never overwritten by content updates from teachers.
- Progress records: Merge strategy takes the highest completion percentage between conflicting versions.
- Notifications: Queued locally, dispatched when connectivity returns.
The sync matrix was the hardest design problem and the most important one to get right before writing a line of sync code. We documented it as a spreadsheet before touching the schema.
AI content pipeline with Human-in-the-Loop review
EduCore includes an AI content generation layer for educators — generating lesson summaries, quiz questions, and reading comprehension exercises from uploaded source material. But no AI-generated content reaches students without passing through a human review queue.
The HITL pipeline works as follows: an educator uploads a document or enters a topic. The AI generates structured content (lesson text, multiple choice questions, grading rubric). The output lands in a review queue visible to the educator and, optionally, a curriculum coordinator. The reviewer can approve, edit inline, or reject. Only approved content is published to the course and synced to learner devices.
"The goal was to make AI a productivity multiplier for overwhelmed educators, not a way to remove human judgment from what students actually learn."
Bring-your-own-model AI routing
EduCore uses an OpenAI-compatible AI router, meaning institutions can plug in their own model — whether that's OpenAI, Anthropic, a local Ollama instance, or a Ministry of Education-approved model. This was a direct response to feedback from NGO partners who cannot send student data to third-party model providers due to data sovereignty requirements.
Visual board system
Beyond the standard course-module hierarchy, EduCore includes a visual board system — a kanban-style workspace where educators can organise content spatially. Cards on the board can be lesson modules, video links, assessments, or external resources. Boards sync offline like everything else and work on both tablet and desktop.
Target markets
Corporate training programmes in Nigeria and Ghana (the first commercial vertical), secondary schools and universities in East Africa, and NGO-run learning programmes across the continent. The platform is also being evaluated by two Southeast Asian regional education ministries for pilot deployments in rural secondary schools.