Inspiration
Artificial Intelligence is transforming education, but millions of students still cannot benefit from it because most AI learning platforms require a stable internet connection. In Nepal, where nearly 65% of the population lives in rural areas, unreliable connectivity creates a major barrier to accessing quality educational resources. We wanted to build an AI solution that adapts to the realities of students instead of expecting students to adapt to technology. That vision inspired GyaanAI—an offline AI learning assistant that makes personalized education available anytime, anywhere.
What it does
GyaanAI is an AI-powered offline learning assistant designed for students in Nepal. It enables students to ask questions, understand concepts, and prepare for exams without relying on continuous internet connectivity. Built around Nepal's national curriculum, GyaanAI provides personalized educational support through on-device AI while ensuring privacy, low latency, and accessibility, even in remote regions.
How we built it
We built GyaanAI using Flutter for the mobile application and Django for backend services. The AI capabilities are powered by Google Gemma running locally through LiteRT, enabling efficient on-device inference. SQLite is used for offline data storage, while Riverpod manages application state. When internet connectivity is available, the application can seamlessly utilize backend services without interrupting the learning experience.
Challenges we ran into
Running a large language model efficiently on mobile devices was our biggest challenge. We had to optimize memory usage, improve inference performance, and ensure a smooth user experience across different Android devices. Building a reliable offline-first architecture while maintaining consistent AI responses required extensive testing and optimization. Another challenge was designing an experience that remained intuitive regardless of whether the device was online or offline.
Accomplishments that we're proud of
- Built a fully functional offline AI learning assistant for mobile devices.
- Successfully integrated on-device AI inference using Google Gemma and LiteRT.
- Created a curriculum-focused learning experience tailored for Nepal.
- Developed a hybrid architecture that supports both offline and online learning seamlessly.
- Made AI-powered education more accessible for students in regions with limited internet connectivity.
What we learned
This project taught us that deploying AI on edge devices involves far more than simply integrating a model. We gained hands-on experience with on-device inference, Edge AI optimization, memory management, offline-first application design, hybrid AI architectures, and building scalable educational technology. More importantly, we learned that impactful innovation starts with solving real problems faced by real people.
What's next for GyaanAI
Our vision is to evolve GyaanAI into a comprehensive AI learning platform for Nepal. Future plans include voice-based tutoring, OCR-powered question solving, multilingual support, teacher dashboards, personalized learning analytics, and broader curriculum coverage. We also plan to optimize the application for production deployment and officially release it on the Google Play Store, making AI-powered education accessible to even more students across the country.
Log in or sign up for Devpost to join the conversation.