Does RChilli use custom embeddings, vector spaces, or deep learning models such as BERT-based or specialized transformer models for semantic similarity?

Yes. RChilli uses BERT-based models and vector database/vector-space technology within its semantic enrichment and normalization architecture.

These technologies are used primarily to enrich and normalize recruitment entities such as Job Profiles, Skills, and Degrees. They are not used as an independent resume-to-job similarity engine that directly determines the final candidate ranking.

The enriched entities are then consumed by Search & Match, which evaluates relevance using entity-level matching, taxonomy and ontology relationships, semantic options, configurable weightage, mandatory criteria, and other Search & Match controls.

A simplified processing flow is:

Resume/JD → Parsing → Entity Normalization and AI/Vector Enrichment → Enriched Job Profiles, Skills and Degrees → Search & Match Relevance/Scoring

RChilli documentation also describes Semantic Search as using AI to identify similar Job Profiles and Semantic Degree Search as using an AI model to retrieve related degrees. This supports the use of AI and vector-based processing for semantic enrichment of recruitment concepts within the broader matching architecture.

Relevant Help Documents:

• Semantic Search
• Semantic Degree
• Ontology Search
• Resume Parser

For more information, contact RChilli Support