CrazyGoldFish Information

CrazyGoldFish is AI evaluation infrastructure for education, the memory layer that turns assessments into structured learning signals. Evaluation is the slowest, most uneven step in education, and it has been treated as a feature instead of infrastructure. CGF scores open-ended answers at scale, handwritten, typed, and multimodal, and captures the reasoning behind each score as structured signal that persists. Ready-to-use APIs and embeddable UIs across the suite: Evaluate: multimodal subjective assessment, rubric or model-answer based. Personalise: evaluation signals into learning pathways. Generate: curriculum-aligned lesson plans, exams, and content. Observe: classroom signals that improve teaching. Engage: parent and child activity packs. Human-in-the-loop is built into the API, not sold on top: teacher review, structured override, and a full audit trail on every scored answer. Beyond scoring, a deterministic adjudicator can judge a disputed mark against the board's own policy and return the same verdict every time it runs, so accountability is part of the evaluation itself. Validated in a government-school pilot featured in MeitY's IndiaAI Impact Summit 2026 Compendium, and in a Hindi-medium, multi-state study with Educate Girls. Built for India's scale and complexity first, then the world.