Implementation & Architecture

Built by Jonas Persson / Sigma Spark AB to help all Sigma colleagues save time and improve the CV matching process for landing more consulting assignments. This comprehensive solution represents over 85,000 lines of code developed over 4 weeks of full-time dedication.

Current Infrastructure

  • Server: Raspberry Pi 4 (8GB RAM)
  • Basic Queue: NVIDIA RTX 3090 (24GB) - LLama 3.1
  • Business Tier: Cloud AI - LLama 3.3
  • Processing: Multi-tier queue system

Development Stack

  • Backend: Node.js with Express
  • Database: PostgreSQL
  • Frontend: Modern HTML5/CSS3/JavaScript
  • AI Models: 5 different LLMs per tier

System Architecture

Frontend Layer Guided Interface, Job Bank, Admin Dashboard Express API Gateway RESTful APIs, Authentication Queue Manager Premium (Cloud AI) Business (Cloud AI) Professional (Local) Basic (RTX 3090) Free (Cloud) AI Processing LLama 3.1 (Local) LLama 3.3 (Cloud) DeepSeek R1 Mistral 7B Multi-Model Support PostgreSQL DB User Management Job Storage Match History Analytics Data Queue Metrics Output Generation PDF Reports, Email Notifications, Real-time Updates

Development Journey

This application was built as a modular system - a journey to deepen competence in testing and development with the help of AI. The entire solution was implemented using Vibe Coding methodology over 4 weeks of full-time development.

AI Tools Used

VSCode + Roo.Code
DeepSeek R1
WindSurf
GitHub Copilot
Claude Code

Time Savings

Average 4 hours saved
per CV matching
Instant results
Multi-language support

Codebase

85,000+ lines total
29,000 JavaScript
27,000 HTML
9,000 CSS
Modular architecture

Performance

Multi-tier processing with automatic scaling. From instant cloud AI to local GPU processing.

13 Languages

Generate motivation letters in Swedish, English, German, French, Spanish, and 8 more languages.

AI-Undetectable

Creates authentic-looking content that bypasses AI detection systems for genuine applications.

Analytics

Real-time dashboard with queue metrics, job performance, and comprehensive match analytics.

To My Sigma Colleagues

As fellow engineers, you understand the complexity of building a system like this. With over 85,000 lines of code (29,000 JS, 27,000 HTML, 9,000 CSS), multi-tier architecture, real-time processing, and AI integration, this represents a significant engineering achievement. But more importantly, it's built to help you succeed.

Good luck with your consultant matching - may this tool help you land many successful assignments!

- Jonas Persson, Sigma Spark AB