A learning system is not only an engineering problem. Our team runs four lines — education research, AI engineering, curriculum content and product design — and a feature ships only when all four agree.
Designs the learning flow and validates outcomes — deciding whether a feature actually helps learning.
Owns model application, the solving and marking pipelines, hint strategy and accuracy.
Writes the question bank and maps it to syllabuses and mark schemes so every question holds up.
Owns interface, information architecture and usability so teachers and students need no training.
AI finds the gaps and generates material; sequencing and judgement stay with the teacher.
Every AI mark and generated item can be edited by a teacher and written back as improvement signal.
We do not chase trends. Every decision returns to one question: does this move a student’s grade.