ENGINEERING PROJECT
Intelligent CV library and semantic search
From heterogeneous documents to profile search based on extraction, structuring and business relevance.
Discuss your projectContext
Context
CVs arrive in several formats with very different levels of structure and readability. Searching profiles using keywords alone is not enough to connect experience, skills and needs. The system had to process both native and scanned documents and make their information usable in business searches.
Digital Place intervention
Digital Place intervention
Digital Place developed a document pipeline covering ingestion, text extraction and OCR, followed by the normalisation of information useful for search. The work focused on structuring extracted content, creating semantic representations and combining vector search with business filters. Backend services and query interfaces made these processes available within profile search journeys.
Architecture
Architecture
Document processing transforms files into structured, indexed information. The search engine matches requests against these representations; APIs connect the engine to consultation and filtering functions.
Technologies : multilingual embeddings.
Engineering decisions
Engineering decisions
Connect extraction and search.
The relevance of a match depends as much on the quality of the extracted information as on the search algorithm that uses it.
Technologies
Technologies
- Docker
- FastAPI
- PaddleOCR
- Qdrant
- React
- TypeScript
Result & scope
Result & scope
- Document pipeline developed
Support for native and scanned CVs, extraction and data structuring.
- Profile search
Semantic indexing, filters and mechanisms for matching profiles to needs.
Explore