Back to docsRecipe

Recipe: Resume search engine

Build a full-text search engine over your resume corpus using Meridian's embedding pipeline and hybrid keyword + vector retrieval.

1

Ingest

Upload PDF/DOCX resumes. Meridian chunks by section and generates embeddings.

2

Index

Hybrid index combines BM25 sparse vectors with dense embeddings for recall.

3

Search

Natural-language queries return ranked candidates with highlighted matches.

Quick start

npx meridian init resume-search

Scaffolds ingestion pipeline, hybrid index config, and a search UI in one command.

Schema design

Each resume document maps to a structured record with raw text, chunked sections, dense embedding, and extracted metadata (name, skills, experience years). The hybrid index runs over both the full-text and vector spaces.

  • BM25 for exact skill matching ("Python 5+ years")
  • Dense vectors for semantic similarity ("backend engineer" matches "API developer")
  • Reciprocal rank fusion combines both result sets
More recipesDocs home