TalentMatch AI

LLM Resume Screening & Recruitment Platform

LLM-Powered Resume Screening & Talent Matching

An automated, persistent NLP candidate evaluation platform for campus placement drives and recruitment sessions. Built using Cosine Similarity, Skill Gap Analysis, and Generative AI.

Core System Architecture & Workflow

01

Recruitment Session & Mode

Create persistent or temporary screening sessions with independent candidate datasets.

02

Multi-Resume PDF Upload

Upload single or bulk candidate resumes with real-time text parsing & size validation.

03

Vector Embeddings

Compute multi-factor similarity (Semantic TF-IDF + Skill Match + Qualifications).

04

Persistent Session Ranking

New candidate uploads are appended and re-ranked across ALL candidates in the drive.

05

Export & One-Click Cleanup

Export Top N candidates to Excel/CSV and clear temporary screening sessions safely.

Step 1: Job Description & Recruitment Session Setup

Select a pre-built campus job role or create a custom recruitment drive.

Select Pre-Built Job Position

Session & Position Details

Step 2: Multiple Candidate Resumes Upload

Upload 1, 5, 10, 20 or more PDF resumes into the current recruitment session.

Upload PDF Resumes

Drag & Drop Multiple PDF Resumes here

or click to browse from your computer (PDF only)

Selected Resumes: 0 files

No files uploaded yet.

Demo Mode: Load Sample Resumes

Use pre-generated campus resumes representing High, Medium, and Low candidate profiles.

Candidate Screening & Persistent Leaderboard

Select a position and upload resumes to calculate candidate rankings.

Candidate Leaderboard

Session: None
0 Candidates Ranked
Rank Candidate Name Match Score Matching Skills Missing Skills Action
No screening performed yet. Click "1-Click Examiner Demo" or upload files.
Recruiter Actions:

System Accuracy & Evaluation Report

Comparison of model recommendations against manually labeled ground-truth campus placement dataset.

Evaluation Dataset: 8 manually labelled resume-JD pairs across 2 campus job positions.
0.0%
Precision
0.0%
Recall
0.0%
F1-Score
0.0%
Top-1 Ranking Accuracy
0.0%
Top-3 Ranking Accuracy

Confusion Matrix

0 True Positives (TP)
0 False Positives (FP)
0 False Negatives (FN)
0 True Negatives (TN)

Evaluation Methodology

  • Ground Truth Dataset: Manually labeled resume–JD pairs with expert relevance tags.
  • Relevance Threshold: Candidates with overall match score ≥ 50% classified as relevant.
  • Top-1 Accuracy: Checks if ground-truth top candidate is ranked #1 by the model.
  • Top-3 Accuracy: Checks if top ground-truth candidate is included in top-3 predictions.

Benchmark Test Cases Breakdown

Target Position Ground Truth Top Candidate Predicted Top Candidate Top-1 Match Top-3 Match