Robotutor
RoboTutor is an intelligent tutoring system from Carnegie Mellon University that enables offline learning in areas with limited network access. Selected as a $1M Global Learning XPRIZE finalist, the project focuses on teaching basic literacy and numeracy skills through Android tablets.
Core Innovation
- Fully offline functionality on Android tablets
- Local storage of student progress and learning data
- Adaptive algorithms operating without cloud connectivity
- Sync capabilities when network becomes available
Technical Implementation
My work focused on the implementation of Multi-armed bandit on the device-side for limited access conditions:
- Built personalized hybrid local-global learning model
- Implemented offline data storage and analytics
- Developed Multi Armed Bandit algorithm using stored weights from previous sessions
- Created sync protocols for intermittent connectivity
Impact
The system demonstrates effective learning technology deployment in network-constrained environments. Its offline-first design ensures consistent education delivery regardless of connectivity, while maintaining the benefits of AI-driven personalization. The Multi Armed Bandit algorithm personalises learning experience for students by adapting difficulty, speed and repititions based on individual performance.