AI Resume & Cover Letter Architect

AI Resume & Cover Letter Architect is a privacy-first, automated career document generator. Instead of manually editing CVs, this system treats a professional's career history as a centralized NoSQL database (MongoDB).

By leveraging a sequential dual-LLM architecture running entirely locally via Ollama, it ensures that your professional data never leaves your machine.

🧠 Architecture

The system utilizes two specialized local models working in tandem:

1. The Reasoner (Llama 3.1): Analyzes the target Job Description and cross-references it with your MongoDB career database to intelligently select the most relevant experiences, skills, and projects.

2. The Coder (Qwen 2.5 Coder): Takes the filtered, hyper-relevant data and safely generates syntax-perfect LaTeX code.

✨ Features

• Privacy-First: 100% local execution using Ollama by default. Zero data is sent to external cloud AI providers.
• Config-Driven Strategy: AI models are decoupled from the agent logic. You can easily upgrade or swap providers (e.g., from Ollama to Gemini) by modifying the .env file and leveraging the BaseAgent interface.
• Database-Driven: Your career is a database. Update your MongoDB documents once, and generate infinite permutations of resumes.
• Hyper-Targeted: Every resume and cover letter is uniquely tailored to the specific job description provided.
• ATS-Friendly: Outputs high-quality, ATS-optimized PDF documents compiled via LaTeX.
• Multilingual Support (i18n): Native support for generating documents in multiple languages using localized fields in the database.

🛠️ Installation & Usage

Prerequisites

• Ollama installed and running locally.
• Models downloaded: llama3.1 and qwen2.5-coder:14b (or 7b).
• Python 3.10+
• MongoDB instance running (local or cloud).

Setup

# Clone the repository and setup virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install project dependencies
pip install -r requirements.txt

# Set up your environment variables
cp .env.example .env
# Edit .env to add your MongoDB URI and set your preferred AI models.
# You can also configure OLLAMA_NUM_CTX in .env to balance memory usage 
# with context limits (e.g., set to 8192 or 4096).

Run the CLI

You can generate either a resume or a cover letter, and optionally specify the language (defaults to en). You can also specify target parameters for better filenames and processing:

python src/main.py [document_type] [language] --role [role] --company [company] --country [country]

# Examples:
python src/main.py resume en --role software-engineer --company amazon --country germany
python src/main.py cover_letter fr --role data-scientist --company google --country france

The generated LaTeX and the compiled PDF will be saved in a Cloudinary-compatible, language-aware folder structure inside the output/ directory (e.g., output/resumes/en/2026-09-04_14-30-00_software-engineer_amazon_germany_en_resume.pdf).

The result is a hyper-targeted, ATS-friendly PDF resume and cover letter generated in minutes, with zero data sent to external cloud AI providers.