Local inference
EduCore explores an adaptive learning device that can operate offline. Its hardware plan combines a 
Raspberry Pi, solar power, and a quantized 
Phi-3 Mini model.
The constraint is the whole system: model memory, response time, energy use, and the learning interface have to fit on one device. Quantization reduces the memory needed to hold model weights.
Project state
The hardware bill of materials is complete. Device testing and evaluation with students are the next questions for the project.