RhaniA

Context

Recruitment in SMBs still relies on Excel files, emails, and manual CV screening. Existing solutions (traditional ATS) are expensive and don't offer intelligent matching.

The problem

How to automatically match candidates to job offers by combining semantic similarity and explicit business criteria, while keeping the scoring transparent and explainable?

Technical solution

RhaniA is a full-stack recruitment platform with hybrid matching.

Architecture

• Frontend: Next.js 14 (App Router), TypeScript, Tailwind, Radix UI, TanStack Query
• Backend: FastAPI, SQLAlchemy 2 async, PostgreSQL 16 + pgvector
• ML: PyTorch, Sentence-BERT (all-MiniLM-L6-v2), vector embeddings
• AI: Gemini Vision (CV parsing), OpenAI/Anthropic (interview questions)
• Infra: Docker Compose, multi-OS startup scripts

Key features

• AI vision CV parsing (PDF → structured data)
• Hybrid matching: cosine similarity (pgvector) + multi-criteria scoring (skills, languages, seniority, contract type)
• Explainable scoring with per-criterion weighting
• Application management with status pipeline
• Interview scheduling with calendar integration (Google / Microsoft)
• LLM-generated interview questions based on profile
• Multi-company support with invitation system
• Built-in English tests
• ~100 backend tests

Notable technical details

• Vector embeddings stored in database (pgvector) for semantic search
• Hybrid scoring combining vectors and business rules with synonyms
• JWT auth with role-based permission matrix
• Strict Router → Service → Repository architecture