A decentralized offline learning platform that uses Bluetooth Low Energy mesh networking to keep coursework moving — even when there is no internet connection at all.
Nigeria has over 200 universities and polytechnics. A significant percentage of students studying outside major cities have unreliable or prohibitively expensive internet access. During my own studies at KWASU, I watched classmates struggle to download lecture materials, submit assignments, or access e-learning portals — not because the content wasn't there, but because the infrastructure wasn't.
The existing solutions — downloaded PDFs, WhatsApp groups, USB drives passed around — were fragile, unsearchable, and impossible to track for progress. I wanted to build something better.
MeshLearn uses Bluetooth Low Energy (BLE) to create a peer-to-peer content distribution network across a campus or classroom. When one device downloads or receives content, it automatically becomes a node that can serve that content to other nearby devices.
This means a single device with connectivity can bootstrap an entire classroom of learners. Content propagates across the mesh in chunks, verified by hash, so corrupted or incomplete chunks are automatically requested from another peer.
The BLE mesh isn't just a data transfer mechanism. It also enables a local social graph — students can see who else in their mesh has completed a module, forming organic study accountability pairs without any server involvement.
MeshLearn includes a lightweight on-device recommendation engine built with TensorFlow Lite. It runs entirely on the student's phone and learns from their engagement patterns to surface the most relevant content modules next.
Training data never leaves the device. The model is updated via a federated learning mechanism — when two devices exchange content over BLE, they also exchange anonymized gradient updates, allowing the model to improve across the mesh without centralizing personal data.
MeshLearn is currently in closed beta with a cohort of students at KWASU. The core BLE content-sharing and progress-tracking features are stable. Active work is ongoing on:
MeshLearn was my first serious encounter with the constraints of mobile hardware. BLE is finicky — Bluetooth stacks vary wildly across Android manufacturers, power management kills background processes unexpectedly, and the mesh topology needs to adapt dynamically as users move around.
It also taught me that the user experience of "it just works offline" is an extraordinarily hard engineering problem. Every assumption you make about connectivity, state synchronization, and data freshness has to be revisited from scratch.
Looking for institutional partners, investors, and NGOs interested in offline-first EdTech in Africa.