Back to Blog
Case Studies 7 min read

Case Study: How Bharat Minds AI Achieved 10x Faster Question Generation

K

Kopfus Delivery

Delivery Team

March 20, 2026

The Challenge

Bharat Minds AI, an edtech platform, needed to generate thousands of high-quality exam questions from textbook content across multiple subjects and difficulty levels. Manual creation was taking their content team 3-4 weeks per subject.

Our Approach

Phase 1: Content Extraction

We built an intelligent document processing pipeline using OCR (Tesseract + custom models) and NLP to extract structured content from textbook PDFs, including diagrams, tables, and mathematical equations.

Phase 2: Question Generation

Using fine-tuned LLMs with curriculum-aligned prompts, we built a system that generates questions with:

  • Multiple difficulty levels (Easy, Medium, Hard)
  • Bloom's Taxonomy alignment
  • Detailed answer explanations
  • Distractor generation for MCQs
  • Phase 3: Quality Assurance

    An AI-powered QA layer checks for factual accuracy, difficulty calibration, and curriculum alignment before human review.

    Results

    MetricBeforeAfterQuestions per week2002,000+Content team size154 (QA focused)Cost per question₹80₹8Time to market4 weeks3 days

    Key Takeaways

  • 1.Domain expertise in the prompting layer is critical for educational content
  • 2.Human QA remains essential — AI generates, humans validate
  • 3.Iterative prompt refinement with teacher feedback dramatically improved quality
  • The project is now in production, powering question banks across 12 subjects for competitive exam preparation.

    #Case Study#EdTech#LLM