Vidyanetra — AI Exam Paper Generator & Assessment Studio

The Operational Problem

Indian schools and coaching institutions spend 4 to 6 hours manually drafting each board-standard examination paper. Faculty grapple with cognitive difficulty drift across different class sections, inaccurate mathematical and chemical typography, out-of-syllabus surprises, and manual margin formatting overhead for print rooms.

The Architectural Solution

Vidyanetra provides an automated examination studio where teachers configure official board blueprints (CBSE, ICSE, State Boards). In 90 seconds, a deterministic RAG engine analyzes exact textbook chapters, balances Bloom's Taxonomy cognitive weightage (Remembering 46%, Applying 22%, Analysing/Evaluating 32%), inserts verified NCERT citations, and exports camera-ready printable PDFs alongside separate step-by-step teacher marking rubrics.

• ✓Dual Creation Engine: Automated deterministic AI paper generator and manual blueprint builder.
• ✓Bloom's Taxonomy cognitive balance: Programmatic compliance with official CBSE/ICSE difficulty spreads.
• ✓Verified NCERT page citations: Zero out-of-syllabus hallucinations with exact book chapter referencing.
• ✓Deterministic 25:50:25 A4 Print Grid: School crest on left, institution title centered, diagonal watermark.
• ✓Bilingual Exam Engine: Formats questions and options in both English and Hindi (Devanagari script).
• ✓Step-Wise Marking Rubrics: Complete partial credit scoring breakdowns for examiner staff rooms.

Technical Anatomy & Data Flow

Engineered as a hybrid distributed system. The retrieval engine is a Python FastAPI service leveraging Pinecone and Qdrant vector backends. PDFs are parsed and indexed by a derived book_id schema (p{pub}-c{class}-s{subject}-h{chapter}). Media and PDF storage are abstracted through a Java Spring Boot storage microservice interfacing with Cloudflare R2, Google Drive, and AWS S3 with machine-to-machine service key authorization. Mathematical formulas render cleanly using KaTeX/MathLive.

{
  "$schema": "https://vidyanetra.in/schemas/blueprint.v2.json",
  "examination": {
    "standard": "CBSE_CLASS_10",
    "subject": "SCIENCE_086",
    "total_marks": 80,
    "duration_minutes": 180,
    "cognitive_balance": {
      "remembering_understanding": 0.46,
      "applying": 0.22,
      "analysing_evaluating_creating": 0.32
    },
    "sections": [
      { "name": "Section A", "type": "MCQ_AND_ASSERTION_REASON", "count": 20, "marks_per_q": 1 },
      { "name": "Section B", "type": "VERY_SHORT_ANSWER", "count": 6, "marks_per_q": 2 },
      { "name": "Section C", "type": "SHORT_ANSWER", "count": 7, "marks_per_q": 3 },
      { "name": "Section D", "type": "LONG_ANSWER", "count": 3, "marks_per_q": 5 },
      { "name": "Section E", "type": "CASE_BASED_INTEGRATED", "count": 3, "marks_per_q": 4 }
    ],
    "citation_policy": "NCERT_TEXTBOOK_2025_EDITION",
    "output_formatting": "A4_PRINT_READY_WITH_MARKING_RUBRIC"
  }
}

Core Technology Stack

Security, Governance & Hardening

Implements token-bucket rate limiting on Gemini AI endpoints, Firebase token validation on every API request, SHA-256 HMAC cryptographic signature verification for Razorpay payment webhooks, and strict role-based access control (Admin, Examiner, Teacher) across institutional departments.