Students use AI to skip thinking. Fermi's smart canvas flips that: students solve by hand while the AI watches every stroke, guiding without solving. The teacher app adds full visibility across classes, homework, tests, and analytics.
MVP & Beta Testing
Beta User Research with Partner Schools
Rigid AI Scoring
Every assignment was scored the same way. Teachers couldn't shape what the AI valued, and students only saw a number with no context for what it meant.
5% steppers gave teachers enough precision to express what mattered, without the overhead of getting it exactly right.
Scaling Personalization
In a class of forty, every student was in a different place. Personalizing homework meant picking questions by hand, so most teachers gave everyone the same worksheet.
Personalised homework, from assignment to assessment
Direction 3 matched how teachers already think: one student at a time. Fermi auto-filled PDF gaps from its question bank, shifting the work from teacher to platform. Grouping by level hid individual differences; topic pooling demanded too much selection. A per-student view surfaced what teachers already tracked in their heads.
Qualitative outcomes from pilot study
Personalised Homework Piloted with 50 students and 4 teachers across two US schools. Instead of one shared worksheet, each student got homework targeted to their weak spots. Fermi's AI flagged the gaps; teachers shaped what came next.
Custom Scoring Rubric Teachers set the grading rules: solving method, steps shown, partial credit. They trusted scores because the rules were theirs. Students trusted them because the calculation was visible. The AI applied the standards consistently.