AI Resume Matcher / RAG System

Overview

AI Resume Matcher is a Spring Boot + Spring AI application that uses Retrieval-Augmented Generation (RAG) concepts to rank resumes against a job description. It parses PDF resumes, chunks content intelligently, generates embeddings, stores vectors in ChromaDB, and returns the best matching candidates with scores and reasoning.

Features

• Upload PDF resumes
• Extract text using Apache PDFBox
• Smart section-based chunking (Summary, Skills, Experience, Projects, Education)
• Local embeddings for semantic search
• ChromaDB vector storage
• Multi-resume ranking engine
• Match score (0-100)
• Matched skills detection
• Relevant excerpts from resumes
• Human-readable reasoning

Tech Stack

• Java 21
• Spring Boot 3.x
• Spring AI
• ChromaDB (Docker)
• Apache PDFBox
• Maven

Architecture

1. Upload Resume PDF
2. Extract Text
3. Chunk Resume into sections
4. Generate Embeddings
5. Store in ChromaDB
6. Accept Job Description
7. Semantic Search + Ranking
8. Return Top Matches

API Endpoints

1. Upload & Embed Resume

POST /api/resume/embed

• Content-Type: multipart/form-data
• Field: file

Response:

Stored 7 chunks for John Anderson

2. Search Resume Chunks

GET /api/resume/search?query=java spring boot

3. Match Job Description

POST /api/job/multi-match

{
  "jobDescription": "Need Java developer with Spring Boot, SQL and REST API experience"
}

Sample Response:

{
  "jobDescription": "Need Java developer with Spring Boot, SQL and REST API experience",
  "topMatches": [
    {
      "candidateName": "John Anderson",
      "resumePath": "resumes/John_Anderson_Resume.pdf",
      "matchScore": 92,
      "matchedSkills": ["Java", "Spring Boot", "SQL"],
      "relevantExcerpts": ["Built scalable microservices..."],
      "reasoning": "Strong match based on semantic relevance and matched skills."
    }
  ]
}

Local Setup

Prerequisites

• Java 21+
• Maven
• Docker

Start ChromaDB

docker run -d --name chroma -p 8000:8000 chromadb/chroma

Learning Outcomes

• Document chunking and embeddings
• Vector databases
• Semantic search
• AI-powered ranking systems
• Spring AI integration

Future Improvements

• Must-have filters (5+ years Java)
• Better scoring formula
• File-system batch loader
• Metrics dashboard
• Frontend UI
• Authentication & role-based access