Back to Projects
Completed Project

CVCraft v2 – AI-Powered ATS Resume Builder

Full-stack ATS resume builder with live editing, real-time ATS scoring with keyword optimization, NVIDIA AI content refinement, instant PDF export, and flexible multi-template MongoDB schema.

|
CVCraft v2 – AI-Powered ATS Resume Builder
RoleFull Stack & AI Engineer
TimelineMay 2026 - Jun 2026
TeamPersonal Project
Tech Stack11 Technologies

Mission Brief

CVCraft v2 is an advanced AI-powered ATS resume builder engineered to help job seekers bypass Applicant Tracking Systems. It delivers real-time ATS compatibility scoring, automated keyword gap analysis, and one-click bullet point refinement powered by NVIDIA NIM (Gemma-2, nv-embed-v1, and Mistral reranker). Built on React 18 and FastAPI with MongoDB persistence, it features instant client-side PDF export (jsPDF), LaTeX server-side compilation, and 84 property-based tests verifying total system integrity — completely free with no signup required.

Key Features

NVIDIA NIM AI Intelligence

  • Gemma-2-2b-it Refinement: Context-aware enhancement of resume work experience bullet points and summaries
  • nv-embed-v1 Semantic Vectors: High-dimension text embeddings analyzing semantic overlap between resume and target job descriptions
  • Mistral Reranking: Precision keyword matching and missing technical terminology suggestions

Live Builder & PDF Engine

  • Live Split Preview: Dynamic real-time rendering updating CV output instantly as user types
  • 2 ATS-Optimized Templates: Jake Ryan Classic Professional and Modern Clean ATS layouts
  • Multi-Tier PDF Export: Instant client-side jsPDF downloads and server-side pdflatex compiling
  • No Signup Barrier: Instant access without authentication barriers or paywalls

Testing & Quality Assurance

  • 84 Property-Based Tests: 42 backend tests via Hypothesis and 42 frontend tests via fast-check
  • Invariant Verification: Score bounds validation, cache deduplication, and fallback behavior when AI tokens deplete
  • SEO & Analytics: Google Analytics 4 tracking with structured JSON-LD and OpenGraph cards

Engineering Chronicles

ATS scoring algorithms often returned erratic match percentages when comparing disparate terminology across industries.

SolutionCombined vector cosine similarity (via NVIDIA nv-embed-v1) with a cross-encoder reranking model (rerank-qa-mistral-4b), stabilizing accuracy across technical domains.

Third-party AI API outages or rate limits could block users from editing and downloading their resumes.

SolutionEngineered graceful fallback mechanisms with local regex-based keyword extractors and in-memory Redis caching, keeping the builder fully operational 100% of the time.

Installation Instructions

1. Clone and Install Dependencies
$git clone https://github.com/KartikSharma4448/CVCraft.git
$cd CVCraft
2. Backend Setup (FastAPI)
$cd backend
$pip install -r requirements.txt
$uvicorn main:app --reload --port 8000
3. Frontend Setup (React)
$cd frontend
$npm install
$npm start

Project Access

Live DemoSource Code

Technologies

FastAPI
React 18
MongoDB 6.0
NVIDIA NIM (Gemma-2-2b-it)
nv-embed-v1
rerank-qa-mistral-4b
Tailwind CSS 3.4
Python 3.10+
jsPDF
Hypothesis
fast-check

Table of Contents

  • • Mission Brief
  • • Key Features
  • • Engineering Chronicles
  • • Visual Gallery
  • • Installation Instructions
Back to Projects

More Projects

PRANAG AI – Web Platform
Completed

PRANAG AI – Web Platform

React.js • AI & Machine Learning

AI Livestock Diagnostic Mobile App
Completed

AI Livestock Diagnostic Mobile App

Flutter • Mobile & On-Device AI

VCC ERP – Coaching Institute Management
Completed

VCC ERP – Coaching Institute Management

Flutter 3.41 • Full Stack & ERP Ecosystem

Rajasthali – Travel & Fleet Management System
Completed

Rajasthali – Travel & Fleet Management System

Flutter 3.7+ • Fleet & Mobile Logistics

Aegis Care – Blockchain Healthcare System
Completed

Aegis Care – Blockchain Healthcare System

Algorand • Web3 & Blockchain