How the AI Works

काम CODE understands meaning — not just keywords. Here's the pipeline that turns your query into the right occupation code.

1. Step 1

   User Search

   A natural language query in English, Hindi or a regional language — typed or spoken.
2. Step 2

   Text Processing

   Normalisation, tokenisation and language detection prepare the query for the model.
3. Step 3

   Embedding Model

   Sentence Transformers / IndicBERT / LaBSE produce a dense vector representing meaning.
4. Step 4

   Vector Database

   FAISS stores embeddings for every NCO-2015 occupation for millisecond nearest-neighbour lookups.
5. Step 5

   Semantic Search & Ranking

   Cosine similarity plus lexical and category boosts rank the best-matching codes.
6. Step 6

   Best NCO Match

   The top-ranked occupations are returned with an AI confidence score and full metadata.