Jodhpur's AI-Assisted Census Assessment — Grading 70,000 Students' Answer Sheets in Days

Manual grading of subjective answer sheets across large public-school systems is slow, inconsistent between markers, and rarely produces topic-level diagnostic data for teachers. Jodhpur district, Rajasthan, piloted an AI-assisted alternative through its Competency-based Census Assessment and School Reporting Pilot Project, run by the district administration and Education Department with technology partners EdOptimize and Central Electronics Ltd (CEL).

To speed up grading of subjective and objective answer sheets and give teachers topic-level competency breakdowns and remedial-intervention suggestions, rather than a single aggregate mark.

Phase 1 (October 2025) covered 54 schools and just over 3,000 students in classes VI-VIII. Phase 2 (April 2026) scaled to more than 1,000 schools across all 15 district blocks and over 70,000 students in classes VI-IX (300,000+ individual evaluations), covering English, Hindi, Mathematics, Science and Social Science in both Hindi- and English-medium schools. Teachers scan completed answer sheets with an AI-enabled application, which grades objective and subjective responses and produces a personalised report card scored 0-3 stars per topic within three days.

Public reporting describes a reduction in grading and reporting turnaround from multiple weeks to three days, and scale-up from 54 schools/~3,000 students to 1,000+ schools/70,000+ students within about six months; however, no independent evaluation of grading accuracy against human markers or of downstream learning-outcome effects has been published.

The initiative is presented as a process-efficiency and diagnostic-reporting success at scale, but the underlying AI model, grading methodology, data-protection and bias-audit details remain undisclosed.