Recruit-AI is an end-to-end agentic resume screening and response system. A recruiter pastes a job description and uploads a candidate resume PDF. Within 10 seconds they receive a full evaluation — match score, executive summary, strengths, gaps, recommendation, and a ready-to-send email draft. Every evaluation is logged in a structured database and visible on a live history dashboard.

Four Layers of the System

• Layer 1 — Capture: Webhook-triggered entry point receives the job description text and resume PDF encoded as base64. A validation node cleans and structures the input before any AI processing begins.
• Layer 2 — Processing: A binary conversion node and PDF extraction node transform the raw base64 into clean readable text regardless of resume layout or formatting.
• Layer 3 — Intelligence: The Gemini agent evaluates the resume against the job description across multiple dimensions and returns a structured JSON evaluation with match score, summary, strengths, gaps, recommendation, and email draft in a single pass.
• Layer 4 — Persistence and Visibility: Every evaluation is logged to Supabase with full breakdown and timestamp. The React frontend surfaces the result immediately and the history page shows all past evaluations with search and filter.

The Full Flow in One Line

JD Text + Resume PDF → Webhook → Validate → Extract PDF Text → Gemini Evaluation → Parse → Supabase → Dashboard

Technical Stack

Average end-to-end evaluation time: Under 10 seconds including database write.