AI CV Customization & Job Automation System

▶ Watch on YouTube

This repository contains a full-stack application for tailoring CVs with AI, analyzing job descriptions, generating cover letters and connection messages, and automating LinkedIn job workflows.

It currently includes:

• A FastAPI backend for CV editing, PDF rendering, job analysis, and browser automation
• A React + Vite frontend for editing resumes, reviewing job matches, and generating content
• Playwright-based LinkedIn automation
• Local LLM inference through Ollama (Qwen 3.5)
• SQLite-backed storage for jobs and LLM evaluation data

---

Tech Stack

Backend

• FastAPI (Python) – API layer and request handling
• SQLite – Lightweight database for jobs and evaluation data
• Playwright – Browser automation for LinkedIn workflows
• Pydantic – Validation and schema enforcement

Frontend

• React – Component-based UI development
• Bootstrap – Styling and responsive layout
• Vite – Frontend build and dev server

AI / LLM Layer

• Ollama – Local LLM runtime
• Qwen 3.5 – Local inference model for CV and cover-letter generation

Infrastructure & Tooling

• UV – Python dependency and environment management
• Node.js + npm – Frontend package management
• Pytest – Backend test suite

---

Table of Contents

• Feature Showcase
• Project Structure
• Architecture Diagram
• Setup
• Ollama Setup
• Running the App
• Testing
• Additional Information

---

Feature Showcase

Job Listings View

---

Resume Builder

---

Connection Message Writer

---

Cover Letter Generator

---

Job Analysis Dashboard

---

Project Structure

AI_CV_Customization/
├── cv_editing_react_app/
│   ├── backend/
│   │   ├── api/                      # FastAPI routes and app entrypoint
│   │   ├── schema/                   # Request/response validation schemas
│   │   ├── services/                 # Business logic, LLM integration, PDF rendering, Playwright automation
│   │   │   ├── browser_automation_playwright/
│   │   │   ├── database/
│   │   │   └── llm.py
│   │   └── test/                     # Backend tests
│   └── frontend/my-app/              # React + Vite frontend app
│       ├── src/                      # React components and pages
│       ├── public/                   # Static assets and HTML templates
│       └── package.json              # Frontend dependencies and scripts
├── job_listings/                      # Saved job listings data
├── package.json                       # Root scripts for running backend/frontend together
├── pyproject.toml                     # Python dependencies and project config
└── assets/                            # Project screenshots and media

The app is now organized under the cv_editing_react_app directory, with the backend and frontend split into separate subprojects.

---

Architecture Diagram

Even a simple text diagram helps:

Frontend (React)
      ↓
FastAPI Backend
      ↓
Playwright Automation → LinkedIn
      ↓
Database (SQLite)
      ↓
Ollama (Local LLM)

Setup

1. Clone the Repository

git clone git@github.com:mohan696matlab/AI_CV_Customization.git
cd AI_CV_Customization

---

2. Install UV and Python Dependencies

If you do not already have UV installed:

pip install uv

Verify the installation:

uv --version

Install the project dependencies and create the virtual environment from the repository root:

uv sync

---

3. Install Playwright Browser Binaries

Playwright is used for LinkedIn automation. Install the browser binaries with:

uv run --directory cv_editing_react_app/backend playwright install

---

4. Configure LinkedIn Credentials

Create a .env file inside the backend directory:

cv_editing_react_app/backend/.env

Add your LinkedIn credentials:

username=your_username
password=your_password

Example:

username=john.doe@gmail.com
password=my_password

---

5. Initialize the Database

Run the database bootstrap from the repository root or from the backend directory:

uv run --directory cv_editing_react_app/backend python -m services.database.db

To reset the database and recreate its tables:

uv run --directory cv_editing_react_app/backend python -m services.database.db --reset

---

6. Start the Backend Server

You can start the backend from the repository root with the helper script:

npm run backend

Or start it directly:

uv run --directory cv_editing_react_app/backend uvicorn api.server:app --reload --port 8000

The API will be available at:

• Application: http://localhost:8000
• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

---

Ollama Setup

Ollama is used for local LLM inference.

1. Install Ollama

Download and install Ollama from:

https://ollama.com/download

Verify the installation:

ollama --version

2. Pull a Model

ollama pull qwen3.5:9b

3. Adjust the Model Settings

The backend currently reads its model configuration from the LLM service file at cv_editing_react_app/backend/services/llm.py. Update the values there if needed:

MODEL_NAME = "qwen3.5:9b"
HOST_ADDRESS = "http://localhost:11434"

4. Remove a Model (Optional)

ollama rm qwen3.5:9b

---

Running the App

Frontend

Make sure Node.js and npm are installed:

node -v
npm -v

From the repository root, start the frontend with:

npm run frontend

You can also start it directly from the frontend folder:

cd cv_editing_react_app/frontend/my-app
npm ci
npm run dev

Full Stack

From the repository root, run both services together:

npm run dev

The frontend will be available at:

• Application: http://localhost:5173/

The backend will be available at:

• Application: http://localhost:8000
• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

---

Testing

Run the backend test suite from the repository root:

uv run pytest -q

---

Docker

Docker Compose runs the React frontend behind Nginx and the FastAPI backend as separate services. SQLite, generated PDFs, and exported job listings are stored in named volumes.

Rotate any credentials that have previously been committed or shared, then create a local Docker env file:

Copy-Item .env.example .env

Fill in GEMINI_API_KEY and APIFY_API_KEY. For LinkedIn automation, also fill in LINKEDIN_USERNAME and LINKEDIN_PASSWORD.

Build and start the default Gemini-backed stack:

docker compose up --build

Open the application at http://localhost:8080. The backend API and Swagger UI are available through http://localhost:8080/api/docs.

To use the optional Ollama container, set MODEL_NAME to the model you pulled and start the profile:

docker compose --profile ollama up --build
docker compose --profile ollama exec ollama ollama pull qwen3.5:9b

Stop the services with Ctrl+C, or use docker compose down. Do not add -v unless you intentionally want to delete the persisted SQLite database and generated files.

---

Useful Commands

Reinstall Project Dependencies

uv sync

Update Dependencies

uv lock --upgrade
uv sync

Install Playwright Browser Binaries

uv run --directory cv_editing_react_app/backend playwright install

Start the Backend Server

npm run backend

Run the Frontend

npm run frontend

Run the Full Stack

npm run dev

---

Additional Information

1. Save LinkedIn Login Cookies (One-Time Setup)

This saves your LinkedIn login session so you do not need to authenticate every time.

uv run --directory cv_editing_react_app/backend python -m services.browser_automation_playwright.save_login_cookies

2. Scrape LinkedIn Job Listings (Optional)

uv run --directory cv_editing_react_app/backend python -m services.browser_automation_playwright.browser_automation

3. Database Management

Remove All Job Records (Keep Database)

uv run --directory cv_editing_react_app/backend python -c "from services.database.db import delete_all_jobs; delete_all_jobs()"

Remove the Database File Entirely

uv run --directory cv_editing_react_app/backend python -c "from services.database.db import remove_db; remove_db()"

---