AI-Powered Candidate Screening System

Overview

AI-Powered Candidate Screening System is a full-stack application that helps automate technical candidate screening using Generative AI and Retrieval-Augmented Generation (RAG).

The system uploads and analyzes a candidate resume, extracts technical skills, lets the candidate select a target role and experience level, generates role-specific interview questions, asks adaptive follow-up questions based on answers, and produces an AI evaluation report.

---

Key Features

Candidate Resume Processing

• Upload PDF resumes
• Extract resume text
• Detect technical skills using an LLM
• Store candidate profile, resume text, and extracted skills

Role-Based Screening

• Fetch available roles from the backend
• Select the role the candidate is applying for
• Match interview context with role requirements
• Support role creation and role deletion through backend APIs

Experience-Level Interviews

• Select candidate experience level before starting the interview
• Supported levels:
  • Fresher
  • Associate
  • Senior
• Send experience_level with the interview start request
• Use experience level as part of the interview session setup

AI Interview Flow

• Generate technical interview questions using RAG
• Show one question at a time
• Ask a follow-up question based on the candidate's answer
• Continue with the next main question after the follow-up
• Store answered questions during the session

AI Evaluation Report

• Generate a final report after the interview
• Summarize candidate performance
• Identify strengths, weaknesses, and knowledge gaps
• Provide AI-generated evaluation feedback
• View a structured recruiter-ready report in the app
• Download the report as a PDF for sharing or record-keeping

---

Technology Stack

Frontend

• React
• Vite
• React Router
• Axios
• Tailwind CSS

Backend

• FastAPI
• Python
• SQLAlchemy
• SQLite
• Uvicorn
• ReportLab
• uv

AI / Machine Learning

• LangChain
• FAISS
• HuggingFace embeddings
• Groq LLM
• Retrieval-Augmented Generation

---

Project Structure

Candidate-Screening-System/
|-- backend/
|   |-- app/
|   |   |-- api/
|   |   |-- database/
|   |   |-- rag/
|   |   |-- schema/
|   |   |-- services/
|   |   `-- main.py
|   |-- books/
|   |-- faiss_index/
|   |-- uploads/
|   |-- interview.db
|   |-- pyproject.toml
|   `-- uv.lock
|
|-- frontend/
|   |-- src/
|   |   |-- components/
|   |   |-- context/
|   |   |-- pages/
|   |   |-- services/
|   |   `-- App.jsx
|   |-- package.json
|   `-- vite.config.js
|
`-- README.md

---

Application Workflow

1. Candidate uploads a PDF resume.
2. Backend extracts resume text and skills.
3. Candidate selects a job role.
4. Candidate selects an experience level: Fresher, Associate, or Senior.
5. Frontend starts an interview session with candidate id, role id, and experience level.
6. Backend generates clean role-specific questions using RAG.
7. Frontend displays one main question at a time.
8. Candidate answers the question.
9. Backend generates a follow-up question based on the answer.
10. Flow continues with the next main question.
11. Candidate ends the interview.
12. Backend generates the final AI evaluation report.
13. Frontend displays a structured evaluation summary.
14. Candidate or recruiter can download the report as a PDF.

---

API Endpoints

Candidate

Method	Endpoint	Description
POST	/upload_resume	Upload candidate resume and extract skills

Role

Method	Endpoint	Description
POST	/create_role	Create a new role
GET	/get_roles	Fetch all available roles
DELETE	/roles	Delete all roles

Interview

Method	Endpoint	Description
POST	/start-interview	Start an interview session
GET	/generate-question/{session_id}	Generate main interview questions
POST	/next-question	Save an answer and generate a follow-up question
GET	/interview/{session_id}	Fetch interview Q&A history
POST	/end-interview/{session_id}	Mark interview as completed
GET	/session/{session_id}	Fetch interview session details
GET	/report/{session_id}	Generate structured interview report
GET	/report/{session_id}/pdf	Download interview report as PDF

Start Interview Request

{
  "candidate_id": 1,
  "role_id": 1,
  "experience_level": "Fresher"
}

Valid experience_level values:

• Fresher
• Associate
• Senior

---

Database Tables

Candidate

Field	Type
id	Integer
name	String
email	String
resume_path	String
resume_text	Text
skills	Text

Role

Field	Type
id	Integer
role_name	String
required_skills	Text

InterviewSession

Field	Type
id	Integer
candidate_id	Foreign Key
role_id	Foreign Key
experience_level	String
status	String

QuestionAnswer

Field	Type
id	Integer
session_id	Foreign Key
question	Text
answer	Text

---

Knowledge Base

The RAG pipeline retrieves interview context from the local knowledge base and vector index. Example resources include:

• Machine Learning by Tom Mitchell
• The Hundred-Page Machine Learning Book
• Machine Learning for Absolute Beginners
• Introduction to Machine Learning with Python
• Master Machine Learning Algorithms
• Pattern Recognition and Machine Learning
• Artificial Intelligence, Machine Learning and Deep Learning

---

RAG Pipeline

PDF documents
    |
    v
Document loading
    |
    v
Text chunking
    |
    v
Embedding generation
    |
    v
FAISS vector store
    |
    v
Semantic retrieval
    |
    v
Question generation

---

Backend Setup

Go to the backend folder:

cd backend

Create and install the environment with uv:

uv sync

Run the FastAPI backend:

uv run uvicorn app.main:app --reload

Backend URL:

http://127.0.0.1:8000

Swagger documentation:

http://127.0.0.1:8000/docs

If port 8000 is already in use, stop the existing process or run on another port:

uv run uvicorn app.main:app --reload --port 8001

---

Frontend Setup

Go to the frontend folder:

cd frontend

Install dependencies:

npm install

Run the frontend:

npm run dev

Frontend URL:

http://127.0.0.1:5173

The frontend uses the Vite dev proxy for backend API calls and does not require backend API URL changes.

---

Environment Notes

The backend uses external AI services. Add required API keys in the backend .env file.

Example:

GROQ_API_KEY=your_api_key_here

HuggingFace may show a warning when running without HF_TOKEN. The app can still run, but setting a token may improve rate limits and download reliability.

---

Report PDF Export

After an interview is completed, the report page includes a Download PDF Report button.

The PDF includes:

• Candidate information (name, email, role, experience level, session ID, status)
• Interview evaluation (overall rating, job eligibility, strengths, weaknesses, knowledge gaps, recommendations, and reason)

The downloaded file is named in this format:

CandidateName_Session_{session_id}_Report.pdf

Example:

Arush_Session_1_Report.pdf

---

Current Limitations

• No separate admin frontend panel yet
• Role creation and role deletion are available through backend APIs only

---

Future Improvements

• Admin dashboard for role management
• Candidate scoring system
• Interview analytics dashboard
• Multi-round interviews
• Authentication for admin and candidate flows

---

Author

Subham