Tech Stack
Python
FastAPI
React
Typescript
PostgreSQL
Docker
Description
ParseHire turns messy resume PDFs into clean, structured candidate profiles. A layout-aware extraction step preserves section structure before an LLM pass maps content into a strict JSON schema validated server-side.
Extracted skills are normalized against a canonical taxonomy (e.g., 'ReactJS', 'react.js' → 'React'), enabling reliable candidate-to-job matching with embedding-based similarity plus rule-based must-have filters.
- Achieved 96% field-level extraction accuracy on a 500-resume evaluation set.
- Schema-constrained LLM output with automatic retry-on-invalid cut malformed parses to under 1%.
- Skill normalization taxonomy of 1,200+ aliases mapped to 300 canonical skills.
- Match scoring combines embedding similarity with hard requirement filters for explainable rankings.
Page Info
Upload & Parsed Profile
Drag-and-drop resume upload with side-by-side view of the original PDF and the extracted structured profile.
Match Scoring
Candidate-to-job-description match scores with per-skill gap breakdown.
