June 2026Hackathon Challenge

CVBlitz — AI Recruitment Platform

A high-fidelity AI-powered candidate ranking platform and recruiter workspace that solves resume fraud and inaccurate skill matching through automated timeline verification, a semantic concept knowledge graph, and explainable AI verdicts at scale. Built for the Redrob AI Hackathon Challenge.

Live Demo View on GitHub

Key Highlights

Sub-4s CPU Processing

Evaluates massive candidate pools on standard CPU within ~3.8 seconds with less than 15MB RAM footprint using native streaming pipelines.

Honeypot Fraud Shield

Identifies impossible chronological overlaps (>2 months) and technology anachronisms (e.g., claiming tools years before their release) to eliminate fake profiles.

Semantic Knowledge Graph

Zero-dependency concept engine leveraging stemming, hypernyms, and domain-aware skill taxonomies for deep contextual match beyond naive keywords.

Features & Capabilities

Recruiter Intelligence Workspace

Gemini JD Parser: Automatically extracts required technical competencies, domain terminology, and role requirements from raw job descriptions via Gemini 2.5 Flash.
Dynamic Multi-Signal Weighting: Real-time slider controls allowing recruiters to balance technical skill fit, behavioral signals, growth trajectory, and authenticity.
Interactive Leaderboard: Rank-ordered candidate roster with tags like Top Match, Hidden Gem, or Honeypot Flagged.
Side-by-Side Radar Visualizer: Compare 2 to 4 candidates simultaneously across technical depth, tenure, growth, and team relevance metrics.
AI Dossier & Explainability: Granular sliding drawers showing evidence-backed recruiter verdicts justifying why a candidate was ranked and scored.

Core Ranking Pipeline

Streamed Ingestion: Zero-dependency streaming ingestion for zipped JSONL files without loading entire datasets into memory.
Multi-Factor Scoring Engine: Computes Technical Fit (40%), Role Relevance (25%), Behavioral Engagement (15%), Career Growth (10%), and Profile Authenticity (10%).
Offline Evaluation Fallback: Seamless fallback mode with pre-mapped challenge datasets when running in air-gapped or API-free environments.

Technology Stack

Frontend Workspace

React 19
TypeScript
TanStack Start & Router
Tailwind CSS

AI & Ranking Engine

Python 3.9+
Google Gemini 2.5 Flash API
Concept Knowledge Graph
Natural Language Processing

Infrastructure & Tools

Vercel Deployment
Vite Build Tool
Lucide Icons & Recharts

Setup & Installation

1. Clone & Install Dependencies

git clone https://github.com/Krisanth-21/CVBlitz.git
cd CVBlitz/talent-geist-ai-main
npm install

2. Configure Environment

# Optional: Provide Gemini API Key (or rely on offline fallback mode)
cp .env.example .env
VITE_GEMINI_API_KEY=your_gemini_api_key_here

3. Run Development Server

npm run dev

4. Python Ranking Pipeline (Optional)

# Run core CPU ranking engine
pip install -r requirements.txt
python rank.py --candidates candidates.jsonl --out submission.csv