marijoAI

This repo contains the code of the website "https://marijoai.com".

marijoAI is a browser-based, no-code tool that helps SaaS and subscription businesses predict customer churn using AI — entirely on your device. It lets you:

• Upload a CSV of customer data and automatically prepare it for training
• Train a neural network to predict which customers will churn
• Score customers with churn risk levels and export results

All processing happens client-side in your browser. Your customer data never leaves your device — no DPA, no vendor risk review, no data compliance concerns.

Features

• Auto CSV detection (header row and delimiter)
• Integrated data cleaning: remove duplicate rows, impute missing values, normalize features
• Optional validation holdout (~20%) with downloadable validation CSV
• Visual step-by-step workflow: Train → Score (plus a Tutorial page)
• Fixed model architecture for simplicity:
  • Input layer: auto-detected feature count from CSV
  • Hidden layer: 64 neurons, ReLU
  • Output layer: 1 neuron, Sigmoid
• Live training feedback with loss and accuracy per epoch
• Churn risk summary: see how many customers are at risk at a glance
• Download trained model as JSON; reload later to score new customer lists
• Scoring reuses training-time preprocessing (feature order and scaling) saved with the model
• Evaluation metrics and confusion matrix when ground truth is available
• Step-by-step tutorial using the included SaaS customer churn dataset

To use the project without an internet connection

1. Clone or download this repository.
2. Open index.html in a web browser. No server required.

Privacy

• Processing is fully in-browser. Files are not uploaded to any server.
• Close the tab to clear in-memory data. Exported files are saved locally to your machine.

License

This project is licensed under the MIT License. See ./LICENSE.

Acknowledgements

• PapaParse for CSV parsing.
• Churn dataset for the tutorial: https://www.kaggle.com/datasets/muhammadshahidazeem/customer-churn-dataset