How Pratham International Uses Claude to Deliver Personalized Student Feedback at Scale

Pratham International is one of India's largest nonprofit education organizations, with over 30 years of history operating in nearly every Indian state and more than 30 countries. Its educational model has been validated through over 10 randomized controlled trials and recognized by institutions such as MIT's J-PAL and the World Bank. Pratham's mission centers on reaching under-resourced communities and providing millions of children access to quality learning.

The central challenge was the inability of teachers to provide meaningful, individualized feedback to large classes. With teachers managing 60 or more students, assessing individual practice exam responses and offering tailored guidance was simply not feasible. Students in under-resourced schools received little to no personalized feedback—a gap with direct consequences for learning outcomes.

Pratham partnered with Anthropic and deployed Claude via the Claude Platform to automate the generation of personalized assessment feedback. The implementation involved extensive iterative prompt engineering, refined with direct support from Anthropic, to ensure feedback aligned with Bloom's Taxonomy—a pedagogical framework covering cognitive levels from factual recall to complex analytical reasoning. The system processes student answers and produces targeted questions and feedback calibrated to each learner's demonstrated level.

The results of the pilot across 20 schools in India were compelling. Over 1,500 student assessments were completed, grading accuracy improved from 30% to 80% through prompt iteration, and the system achieved 90% accuracy in generating Bloom's Taxonomy-aligned questions. The deployment demonstrates how AI-powered personalization can help under-resourced educational systems bridge the feedback gap at scale.