Skip to main content

Hiring Assistant

Created on April 4|Last edited on April 4


Project Overview

The Hiring Assistant is an AI-powered system designed to automate and improve the hiring process by evaluating job applications against job offers. The system uses advanced language models and a structured workflow to make informed hiring decisions while maintaining fairness and reducing bias.


System Architecture

Core Components

  1. Document Processing
  • PDF text extraction for both job offers and applications
  • Image processing capabilities for visual content
  • Structured data extraction using AI models
2. AI Models Integration
  • Multiple model support (OpenAI, AWS Bedrock, Ollama)
  • Model selection and configuration through UI
  • Guardrail mechanisms to prevent hallucinations
3. Evaluation System
  • Single test mode for individual applications
  • Batch testing for multiple evaluations
  • Comprehensive scoring system

Workflow Structure

The system follows a multi-step workflow:
  1. Document Upload
  2. Information Extraction
  3. Comparison and Decision Making
  4. Hallucination Checking
  5. Expert Review (when needed)



Key Features

Dataset Creation

The system includes a sophisticated dataset creation process:
  1. Characteristic Generation
  • Configurable number of applicants
  • Bias control mechanisms for:
    • Gender
    • Age
    • Nationality
  • Structured applicant profiles
2. R Score Calculation
  • Quantitative measure of dataset representativeness
  • Statistical analysis of distribution
  • Threshold-based validation
3. Application Generation
  • AI-powered application creation
  • Integration with job offer requirements
  • Quality control mechanisms


Evaluation Modes

  1. Single Test Mode
  • Direct comparison of job offer and application
  • Real-time decision making
  • Detailed reasoning output
  • Expert review integration
2. Batch Testing Mode
  • Large-scale evaluation capabilities
  • Multiple trial support
  • Comprehensive scoring system
  • Performance metrics

Expert Review System

  • Human-in-the-loop integration
  • Configurable review triggers
  • Expert override capabilities
  • Decision tracking and logging