LearnLocal

A resilient, offline-first learning app for children that utilises ollama and local GPT-OSS models for AI-powered education. Provides continuous learning access even in challenging circumstances.

Comment

LearnLocal

A resilient, offline-first learning app built with Expo that integrates with FastAPI and local GPT-OSS models for AI-powered education. Designed to provide continuous learning access even in challenging circumstances.

Inspiration

Our team’s motivation comes from lived experience. One of us was born into conflict in Kosovo during the war, while others grew up in the Global South, where lower levels of economic development made access to education uncertain. We’ve all witnessed firsthand how fragile schooling can be when infrastructure or stability breaks down.

More recently, as adults, we also lived through COVID-19 disruptions, watched the wildfires in LA, the conflicts in Middle east which showed us again how vital continuity of education is for wellbeing and hope. These experiences shaped our belief that access to learning should never depend on circumstances.

That’s why we built LearnLocal: to ensure no child’s education is interrupted — not by war, disaster, displacement, or isolation.

What it does

LearnLocal is an offline-first, AI-powered learning app that provides structured lessons, dynamic AI-generated content, and multi-language support. It works fully offline, stores progress locally, and can share content between devices without internet — so one device can educate many children in emergencies.

Core features include:

  • AI chat assistant powered by local GPT-OSS models
  • Dynamic lesson generation on any topic
  • Interactive modules with progress tracking and gamification
  • Multi-language translation and text-to-speech
  • Device-to-device sharing via WiFi Direct, Bluetooth, hotspot, or QR code
  • Crisis-ready deployment for refugee camps, hospitals, and disaster shelters

How we built it

  • Frontend: React Native with Expo SDK for mobile and web, styled with a consistent design system
  • Backend: FastAPI server with RESTful endpoints, plus an embedded server mode for complete offline functionality
  • AI Integration: Lightweight GPT-OSS models running locally for chat, lesson generation, translation, and text-to-speech
  • Offline-first architecture: Smart caching, local storage with AsyncStorage, and network-aware fallback operations
  • Content sharing: Implemented WiFi Direct, Bluetooth Low Energy mesh, hotspot serving, and QR-based transfer
  • Deployment: Cross-platform builds for Android, iOS, web (PWA), and emergency offline kits

Challenges we ran into

  • Optimizing AI model performance on low-resource devices
  • Designing an intuitive offline-first experience that feels seamless to users
  • Ensuring battery efficiency and power management for crisis environments
  • Building a child-friendly UI while handling advanced technical features
  • Integrating multiple device-to-device sharing protocols reliably

Accomplishments that we’re proud of

  • Built a working offline-first education platform that runs with or without internet
  • Proved that one device can serve many children simultaneously in shelters or classrooms
  • Designed with child wellbeing in mind: trauma-informed, culturally adaptable, and multi-language
  • Delivered a scalable architecture that works on mobile, web, and as a Progressive Web App
  • Created a project that reflects our personal mission: making education resilient for children in crisis

What we learned

  • Building offline-first systems means unlearning many assumptions about constant connectivity
  • Humanitarian solutions must prioritize energy efficiency, accessibility, and cultural adaptation
  • Local AI is feasible but requires careful optimization and fallback mechanisms
  • Education technology must integrate pedagogical and psychological insights alongside code

What’s next for LearnLocal

  • Enriching contents with online research agent when you do go online to get latest context aware information
  • Feedback Loop
  • Authentication and Security
  • Partner with NGOs, hospitals, and schools for pilot deployments in real-world crisis zones
  • Expand localized content libraries aligned with regional curricula
  • Add more accessibility features for children with disabilities
  • Further optimize for low-end smartphones and tablets with smaller AI models
  • Build community features like peer learning and group challenges to strengthen social connection

✨ LearnLocal: Ensuring no child’s education is interrupted, no matter the circumstances.

Built With

  • fastapi-(python)
  • gpt-oss
  • react-native-(expo)
  • typescript
Share this project:

Updates