1. Data Ingestion & OCR
Teachers upload raw images of ZIMSEC past papers or handwritten student scripts. Our optical character recognition (OCR) engine extracts the text, mathematical formulas, and diagrams with extremely high fidelity, converting them into structured JSON data.
2. Vectorization & RAG Storage
The extracted text is sliced into semantic chunks and embedded into a high-dimensional vector space using Pinecone. This Retrieval-Augmented Generation (RAG) architecture equips students with instant grounded tutoring, provides teachers with instant lesson resources, and gives headmasters curriculum alignment assurance.
3. Pedagogical Routing
When a query or command is executed, it is intercepted by our Orchestrator Agent. Based on role context (Student, Teacher, or Headmaster), grade level, and objective, the Orchestrator routes tasks to specialized sub-agents—delivering Socratic guidance for students, auto-grading for teachers, and cohort performance telemetry for headmasters.
4. Socratic Generation
The specialized agent synthesizes role-tailored outputs. Students receive strict Socratic guidance that builds critical thinking, teachers receive immediate diagnostic insight into student misconceptions, and institutional headmasters gain real-time visibility into subject mastery across entire departments.