JobPlatform is a full-stack MERN-based recruitment system where:
- HR can post job openings
- Candidates can apply for jobs
- AI automatically shortlists candidates
- An AI Agent handles resume evaluation and candidate interaction
- No human bias
- No manual shortlisting
- Fully automated intelligent screening process
This platform eliminates traditional manual screening and replaces it with a semantic AI-powered evaluation engine.
HR creates a job with:
- Required skills
- Experience requirements
- Role description
Users upload resumes and apply to jobs.
The system automatically:
- Converts Job Description into embedding
- Converts Resume into embedding
- Calculates semantic similarity
- Extracts skills
- Detects experience
- Generates final weighted score
Candidates are shortlisted based on intelligent scoring.
The AI Agent:
- Reads the entire resume
- Understands candidate background
- Conducts automated conversation
- Evaluates responses
- Maintains structured scoring
- Works without human interruption
This ensures:
- Zero bias
- Consistent evaluation
- Fair candidate assessment
- Scalable recruitment process
An AI-powered Resume Ranking system built using Node.js + Transformers that semantically matches resumes with job descriptions.
This system automatically ranks resumes based on:
- Semantic similarity (AI-based understanding)
- Required skills match
- Experience indicators
- Weighted scoring model
It works like a modern AI-powered Applicant Tracking System (ATS).
@xenova/transformers
- JavaScript implementation of HuggingFace Transformers
- Runs locally in Node.js
- No external API required
- No paid AI service needed
Xenova/all-MiniLM-L6-v2
- Lightweight transformer model
- Converts text → 384-dimensional embedding vector
- Optimized for semantic similarity
- Fast and production-ready
- Ideal for backend systems
- Convert Job Description → Embedding vector
- Convert Resume → Embedding vector
- Calculate Cosine Similarity
- Extract required skills from JD
- Calculate skill match score
- Detect experience signals
- Compute final weighted score
Final Score = (50% Semantic Similarity) + (30% Skill Match) + (20% Experience Score)
<<<<<<<<< Temporary merge branch 1
| Traditional (TF-IDF) | This System |
|---|---|
| Keyword matching only | Understands meaning |
| No context awareness | Semantic understanding |
| Weak filtering | Intelligent ranking |
npm install @xenova/transformers
========= Score is returned as a percentage (0–100).
Traditional (TF-IDF) This System Keyword matching only Understands meaning No context awareness Semantic understanding Weak filtering Intelligent ranking
npm install @xenova/transformers
AI-based semantic ranking
Skill extraction from job description
Experience detection
Weighted ATS-style scoring
Fast & scalable
Works offline after first model load
Bias-free candidate evaluation
AI-driven automated shortlisting
Dynamic skill extraction
Years of experience detection
Keyword stuffing detection
Batch embedding optimization
Store embeddings in database for faster ranking
Real-time AI interview scoring
AI behavioral analysis module
Node.js
Express
MongoDB
Transformers (MiniLM)
MERN Stack Architecture
JobPlatform demonstrates how modern AI and transformer-based embeddings can revolutionize recruitment by:
Replacing keyword filtering with semantic understanding
Automating candidate shortlisting
Enabling AI-driven interviews
Removing human bias from early-stage hiring
Creating a scalable, intelligent hiring ecosystem
Temporary merge branch 2