Recruiting is rarely a fixed search problem. Preferences change as teams meet candidates and refine what success looks like. Pariti’s later system was built around that reality.
The pipeline extracted job requirements, normalized CVs, combined embeddings with boolean constraints, and returned a ranked set of candidates. Recruiters could then adjust priorities instead of accepting an opaque score.
The technical record describes the underlying architecture: open-source vectorization, a CV parsing dependency, Qdrant retrieval, structured requirements, and an interactive recruiter-facing layer.