🚀 TalentMind-AI

AI-Powered Intelligent Candidate Discovery System

Built for the Redrob AI Hackathon – Intelligent Candidate Discovery Challenge

---

📌 Overview

TalentMind-AI is an AI-powered recruitment assistant that intelligently discovers and ranks the most relevant candidates for a given Job Description (JD). Instead of relying on traditional keyword-based matching, the system understands the semantic meaning of both job descriptions and candidate profiles using transformer-based embeddings.

The retrieved candidates are further re-ranked using a Hybrid Ranking Algorithm that combines semantic similarity, skill matching, experience, title relevance, and behavioral signals to generate an explainable ranked shortlist.

---

🎯 Problem Statement

Traditional Applicant Tracking Systems (ATS) rely heavily on keyword matching, often missing qualified candidates whose resumes use different terminology.

TalentMind-AI addresses this challenge by:

• Understanding Job Descriptions semantically
• Processing candidate profiles intelligently
• Retrieving relevant candidates using vector similarity search
• Ranking candidates using multiple recruiter-centric signals
• Providing explainable recommendations for recruiters

---

📂 Dataset

The complete dataset used in this project is provided as part of the Redrob AI Hackathon.

Due to GitHub file size limitations, the full dataset is not included in this repository.

Expected dataset location:

data/raw/
    candidates.jsonl
    job_description.docx

✨ Features

• 📄 Job Description Parsing
• 👤 Candidate Profile Processing
• 🧠 Semantic Embedding Generation
• 🔍 FAISS Vector Similarity Search
• 🤖 Hybrid Candidate Ranking
• ✅ Intelligent Skill Matching
• 📈 Explainable Score Breakdown
• 📊 Ranked Excel Report Generation
• ⚡ Fast Candidate Retrieval

---

🧠 AI Pipeline

Job Description (.docx)
        │
        ▼
JD Parsing & Requirement Extraction
        │
        ▼
Sentence Transformer
(all-MiniLM-L6-v2)
        │
        ▼
Semantic Embedding
        │
        ▼
FAISS Vector Search
        │
        ▼
Top-K Candidate Retrieval
        │
        ▼
Hybrid Ranking
(Semantic + Skills + Experience +
Title + Behavioral Signals)
        │
        ▼
Ranked Candidate Recommendations
        │
        ▼
Excel Output

---

🛠 Tech Stack

Category	Technologies
Language	Python 3.10
AI / NLP	Sentence Transformers (all-MiniLM-L6-v2)
Vector Search	FAISS
Document Processing	python-docx
Data Processing	NumPy, Pandas
Excel Export	OpenPyXL
Progress Tracking	tqdm
IDE	VS Code

---

📂 Project Structure

TalentMind-AI/
│
├── data/
│   ├── raw/
│   └── processed/
│
├── models/
│
├── outputs/
│
├── src/
│   ├── embeddings/
│   ├── explainability/
│   ├── jd/
│   ├── output/
│   ├── preprocessing/
│   ├── ranking/
│   ├── retrieval/
│   └── utils/
│
├── main.py
├── requirements.txt
├── README.md
└── recommended_candidates.xlsx

---

⚙️ Installation

Clone the repository

git clone https://github.com/Meghana-539/TalentMind-AI.git

Move into the project

cd TalentMind-AI

Install dependencies

pip install -r requirements.txt

---

▶️ Running the Project

Generate Candidate Embeddings

python -m src.embeddings.generate_embeddings

Build FAISS Index

python -m src.retrieval.build_index

Retrieve & Rank Candidates

python -m src.retrieval.search

Export Ranked Candidates to Excel

python -m src.output.export_excel

---

📊 Sample Output

The system generates:

recommended_candidates.xlsx

The report includes:

• Rank
• Candidate ID
• Current Title
• Company
• Experience
• Top Skills
• Matched Skills
• Semantic Score
• Skill Score
• Behavior Score
• Title Score
• Final Hybrid Score

---

🔍 Explainability

Unlike traditional black-box ranking systems, TalentMind-AI provides transparent recommendations by displaying:

• Semantic Similarity Score
• Skill Match Score
• Experience Score
• Title Relevance Score
• Behavioral Score
• Final Hybrid Score
• Matched Skills

This enables recruiters to understand why each candidate was recommended.

---

🚀 Key Advantages

• Semantic understanding instead of keyword matching
• Efficient retrieval across 100,000 candidate profiles
• Hybrid multi-factor ranking
• Explainable AI recommendations
• Fast FAISS vector search
• Modular and scalable architecture

---

🔮 Future Enhancements

• Streamlit Recruiter Dashboard
• Resume PDF Upload
• LLM-powered Candidate Summary
• Interview Recommendation Engine
• Multi-language Resume Support
• Recruiter Analytics Dashboard
• Real-time Candidate Database Updates

---

👩‍💻 Developer

Meghana Pujari

B.Tech – Robotics & Artificial Intelligence

SASTRA Deemed University

---

🏆 Hackathon

Developed for the Redrob AI Hackathon – Intelligent Candidate Discovery Challenge.

---

📜 License

This project was developed as part of the Redrob AI Hackathon for educational and demonstration purposes.