Recruitt: AI Talent Screening Platform

Platform Preview

Recruitt is an advanced AI-powered platform designed for the Umurava AI Hackathon. It streamlines the talent acquisition process by leveraging Gemini AI to accurately, transparently, and efficiently shortlist job applicants.

Core Features

  • Full Lifecycle Job Management: Create, view, edit, and delete job openings with specialized metadata fields.
  • Smart Ingestion: PDF resume parsing and bulk CSV/XLSX mapping using Gemini AI.
  • AI Screening Matrix: Multi-candidate evaluation based on job-specific requirements.
  • Weighted Scoring: Recruiter-controlled scoring priorities (Skills vs Experience vs Education).
  • Explainable Results: Detailed AI reasoning for every match, including strengths and risks.
  • Human-in-Loop Notifications: Staged email drafts for shortlisted and rejected candidates.

Integrated Architecture

Recruitt follows a modern, full-stack monorepo architecture using Turborepo, ensuring seamless synchronization between the frontend and backend.

Recruitt Integrated Architecture

Strategic Integration Points:

  • Unified Schema: Both layers share the same Talent Profile Specification, preventing data drift.
  • AI Orchestration: The backend manages prompt engineering and context window optimization, while the frontend provides interactive controls for AI weighting.
  • Staged Notifications: A bridge between AI evaluation and human finality, allowing recruiters to review and bulk-send notifications generated by the screening results.

Quick Test Guide

Follow these steps to experience the full power of Recruitt in minutes:

  1. Launch Environment: Run pnpm dev from the root directory.
  2. Define a Role: Create a new job posting with specific skills and requirements.
  3. Bulk Ingestion:
    • Go to your new job's details.
    • Click "Add multiple applicants".
    • Upload the sample file: apps/server/dummy_applicants.csv.
  4. Execute Screening:
    • Click "Screen Candidates".
    • Adjust the Weighting Sliders (e.g., prioritize Experience over Education).
    • Watch the AI rank the candidates in real-time.
  5. Decision Making: Shortlist or Reject candidates to generate automated email drafts in the Notification Center.

Technology Stack

  • Language: TypeScript (Strict mode + Comprehensive JSDoc)
  • Frontend: Next.js 16, Tailwind CSS 4, Shadcn/ui, Tabler Icons
  • Backend: Node.js, Express, Mongoose, Zod
  • AI Engine: Gemini 2.5 Flash
  • Database: MongoDB Atlas

Getting Started

Prerequisites

  • Node.js 20+
  • pnpm 9+
  • Gemini API Key & Resend API Key (for emails)

Installation

# Install dependencies
pnpm install

# Start development environment
pnpm dev
Edit this page on GitHub
Table of Contents