Engineering Smart Assessments with AI Question Intelligence
From raw past paper PDFs to syllabus-aligned mock tests. Powered by Gemini AI question generation, pgvector semantic deduplication, and automated candidate scorecards for institutions.
100%
Automated Generation
Zero
Duplicate Questions
Advanced
Topic Templates
Real-Time
Assessment Scoring
From Raw Data to Real Exams in Minutes
Our intelligent pipeline handles everything from ingesting dusty PDFs to administering high-stakes, time-bound assessments with strict topic distributions.
1. Ingest & Generate
Upload existing PDFs for OCR extraction, or command Google Gemini to generate fresh, syllabus-aligned questions on demand.
2. Build Templates
Create dynamic Topic Percentage Templates to strictly control how many questions from each sub-topic appear in the final exam.
3. Administer Exam
Candidates take the exam in a secure, time-bound environment with interactive question navigation and progress tracking.
4. Analyze Results
Instant grading provides deep analytics, including category-wise performance and granular topic scoring breakdowns.
Under the Hood Architecture
Built on strict business rules ensuring correctness, token management, and data integrity.
Hybrid PDF & OCR Pipeline
PyMuPDF extracts native text instantly. Scanned or low-density pages seamlessly trigger Pillow contrast thresholding and Tesseract OCR with regex question & answer mapping.
Safe Gemini Batch Generation
Generates original MCQs in small 5-question batches with exponential rate-limit backoff, strict JSON schema validation, and human review approval gates.
Vector Duplicate Detection
Dual duplicate engine: Normalized SHA-256 hash for exact matches, combined with PostgreSQL pgvector cosine similarity checks (≥ 0.88 duplicate, ≥ 0.75 review).
Topic Percentage Templates
Dynamically assemble exams by defining min/max percentage targets for different topics. The engine algorithmically allocates exact question counts.
Interactive Assessment Engine
Candidate-facing exam portal with countdown timers, mark-for-review capabilities, and local-storage persistence to prevent data loss during network drops.
Granular Performance Analytics
Post-exam dashboards break down scores by category, highlight strengths and weaknesses across sub-topics, and filter answers by correct/wrong/unanswered.