zivAI is an AI-powered teacher and student assistant designed for O-Level education. Built by Team ZWASCEND at the University of Zimbabwe, it combines a full learning management system with automated marking, student support, and offline-first deployment for schools with limited connectivity.
I worked as team captain and solutions architect for the project. The system was built for the Huawei ICT Innovation Competition 2025-26, where it won Third Prize at the Southern Africa Regional Finals.
Platform Architecture
The platform uses a hybrid cloud-edge model. Schools can run an edge node locally for class management, assessments, submissions, review workflows, reporting, and AI-assisted feedback. When connectivity is available, the edge node synchronizes data back to the cloud.
The LMS backend is built with Spring Boot and PostgreSQL. It handles business logic, security, REST APIs, assessment workflows, and edge-cloud synchronization.
AI Services
The AI layer is exposed through Flask services behind clean APIs. These services support automated short-answer grading, knowledge tracing, and OCR-based marking.
Models were trained on Huawei ModelArts with MindSpore and packaged for offline edge serving with MindSpore Lite.
Frontend And Infrastructure
The frontend is built with React across six repositories. The deployment uses Huawei Cloud services including CCE, ECS, RDS for PostgreSQL, and OBS for storage.
The codebase is available at github.com/zwAscend.