🧠 AI-Powered Multi-Agent Job Screening System

🚀 Problem Statement 5: Enhancing Job Screening with AI and Data Intelligence

Manual job screening is often inefficient, time-consuming, and error-prone. This system aims to automate and optimize the recruitment process using a multi-agent LLM-based architecture and data intelligence techniques.

---

🎯 Objective

Develop a multi-agent AI system that:

• Reads and summarizes Job Descriptions (JDs)
• Extracts structured data from candidate CVs
• Matches qualifications with JDs using role-specific evaluators
• Shortlists candidates
• Sends automated interview invitations

---

🧱 Multi-Agent System Architecture

🟢 Layer 1: Information Extraction Agents

1. JD Summarizer Agent

   • Parses and summarizes job descriptions
   • Extracts key requirements: skills, qualifications, responsibilities

2. CV Extractor Agent

   • Extracts structured data from resumes
   • Key info: education, experience, certifications, soft/technical skills

---

🟡 Layer 2: Role-Based Evaluation Agents

These agents evaluate the match between JD and CV from different professional lenses:

1. HR Agent

   • Assesses communication skills, culture fit, and soft skills
   • Evaluates from a human resource perspective

2. Technical Agent

   • Matches technical stack, programming languages, frameworks
   • Evaluates domain-specific expertise

3. Business Agent

   • Evaluates business fit and general English proficiency
   • Checks alignment with business goals and client interaction needs

Each agent returns a score out of 100 and an explanation for their evaluation.

---

🔵 Layer 3: Final Decision Agent

• Takes input from all three agents (HR, Tech, Business)
• Applies weighted logic to combine the results
  Example:
  Final Score = 0.3 * HR + 0.5 * Tech + 0.2 * Business
• If the final score ≥ 80%, the candidate is shortlisted
• Generates and sends interview invitation emails with scheduling details

---

💾 Tech Stack

Component	Technology
Frontend	React , Tailwind CSS
Backend	NodeJs , ExpressJs
MultiAgents	OpenAi , Langchain , Langgraph , Langamith
DataBase	SQLite
Version Control	Git , Github

---

🗃️ SQLite Database Structure

job_descriptions Table

| id | title | summary | skills_required | created_at |

candidates Table

| id | name | email | resume_text | extracted_skills | match_score | hr_score | tech_score | business_score | final_score | shortlisted (bool) |

interviews Table

| id | candidate_id | date | time | interview_format | email_sent (bool) |

---

⚙️ Flowchart

+-------------------------+
|   Upload JD + CVs      |
+-------------------------+
           |
           v
+-------------------------+
| Layer 1: JD + CV Agents |
+-------------------------+
           |
           v
+-----------------------------+
| Layer 2: Evaluation Agents |
| (HR, Tech, Business)       |
+-----------------------------+
           |
           v
+-------------------------+
| Final Decision Agent    |
+-------------------------+
           |
   If final_score >= 80%
           |
           v
+-------------------------+
| Send Interview Email    |
+-------------------------+

---

📌 Key Features

• ✅ Automated JD & CV analysis
• ✅ Multi-role evaluation
• ✅ Scoring with explainability
• ✅ SQLite-based long-term memory
• ✅ Interview automation

---

🧠 Future Enhancements

• Resume parsing for multiple languages
• Adaptive weights for final decision agent
• Integration with Google Calendar for interview scheduling
• Admin dashboard to review candidate pipeline

---

🤝 Contributing

We welcome contributions! Please fork the repo and create a PR. Let's build smarter hiring together 💼🧠

---

👩‍💻 Team

• Avaniben Kanjibhai Prajapati
• Amankumar RobinBhai Galoliya

---