PYQ Atlas: How we indexed 8,400 NTA questions into one heatmap.

We built a crawler, a tagger, and a frequency analyser. Then we found that 87% of NEET marks come from 40% of chapters. Here's the data.

When you look at 8,400 questions over a decade of exams, patterns start to emerge. The human brain is terrible at visualizing this scale of data, so we built a heatmap.

The Engineering Challenge

Tagging 8,400 questions manually is impossible to do consistently. We used a multi-pass pipeline:

1. First, an OCR pass to extract text and math symbols perfectly.
2. Second, a semantic mapping pass using a custom embedding model fine-tuned on JEE/NEET syllabi.
3. Third, human verification on edge cases.

The result is the PYQ Atlas. Explore it in the app today.