Custora Ai – End-to-End Machine Learning SaaS Application

2026 • Suraj Singh
Data Science

date: 2026-02-19

Logo

Overview

Custora AI is a full-stack Machine Learning web application built using Flask that enables businesses to upload customer datasets and predict churn probability in real time.

This project demonstrates real-world ML deployment from model training to a working SaaS-style web interface.

It is not just a model. It is a complete product pipeline.

Problem Statement

Customer churn directly affects business revenue.
Many companies struggle to analyze churn risk from raw customer data.

Custora AI solves this by:

  • Accepting structured CSV datasets
  • Processing telecom-style customer data
  • Generating churn probability predictions
  • Displaying percentage-based risk scores
  • Helping businesses take proactive retention decisions

Key Features

  • 🔐 User Registration & Login System
  • 📊 Interactive Dashboard
  • 📂 CSV Upload Support
  • 🤖 Machine Learning Model Integration
  • 📈 Churn Probability Output (%)
  • 📥 Downloadable Prediction Results (CSV)
  • 🔎 First 10 Rows Preview for Large Datasets
  • ⚡ Lightweight and Fast Flask Backend

Tech Stack

Backend

  • Python
  • Flask
  • Pandas
  • NumPy
  • Scikit-learn
  • Pickle

Frontend

  • HTML
  • CSS
  • JavaScript
  • Jinja2 Templates

Database

  • SQLite (Flask Instance DB)

Machine Learning Pipeline

  1. Data Cleaning & Preprocessing
  2. Feature Encoding
  3. Model Training (Classification Model)
  4. Model Serialization using Pickle
  5. Model Deployment inside Flask App
  6. Real-time CSV Prediction Handling

📂 Project Structure

├── 📁 .ipynb_checkpoints
│   └── 📄 customer churn data analysis-checkpoint.ipynb
├── 📁 .virtual_documents
│   └── 📄 customer churn data analysis.ipynb
├── 📁 Data
│   ├── 📁 .ipynb_checkpoints
│   │   └── 📄 Telco-Customer-Churn-checkpoint.csv
│   └── 📄 Telco-Customer-Churn.csv
├── 📁 instance
├── 📁 models
│   ├── 🐍 __init__.py
│   └── 🐍 user.py
├── 📁 routes
│   ├── 🐍 __init__.py
│   ├── 🐍 auth.py
│   └── 🐍 main.py
├── 📁 services
│   ├── 🐍 __init__.py
│   ├── 🐍 ai_summary.py
│   ├── 🐍 pdf_report.py
│   └── 🐍 prediction.py
├── 📁 static
│   ├── 📁 css
│   ├── 📁 images
│   │   ├── 🖼️ 4-removebg-preview.png
│   │   ├── 🖼️ dark-logo.png
│   │   └── 🖼️ light-logo.png
│   └── 📁 js
├── 📁 templates
│   ├── 🌐 base.html
│   ├── 🌐 dashboard.html
│   ├── 🌐 landing.html
│   ├── 🌐 login.html
│   ├── 🌐 register.html
│   ├── 🌐 results.html
│   └── 🌐 upload.html
├── 📁 utils
│   ├── 🐍 __init__.py
│   └── 🐍 data_validation.py
├── ⚙️ .gitignore
├── 📄 Procfile
├── 🐍 app.py
├── 📄 churn_model.pkl
├── 🐍 config.py
├── 📄 customer churn data analysis.ipynb
├── 🐍 extensions.py
├── 📝 readme.md
├── ⚙️ render.yaml
└── 📄 requirements.txt

Sample Output

Churn Probability (%) Prediction
96.52 1
2.83 0
72.84 1
8.10 0

Challenges Faced During Development

  • API key configuration issues
  • Handling large CSV uploads
  • Preventing frontend crash during heavy data rendering
  • file structuring
  • Model integration with real-time user uploads

Future Improvements

  • Multi-model support (Modular ML tools)
  • Admin analytics dashboard
  • Deployment on cloud (AWS / Render / Railway)
  • Payment integration for SaaS model
  • Better UI/UX enhancements
  • Role-based access control
  • Model performance monitoring

Learning Outcomes

  • End-to-end ML deployment
  • Full-stack application structuring
  • SaaS-oriented thinking
  • Real-world dataset handling
  • Scalable project architecture design

Author

Suraj Singh Linkedin
Aspiring Data Scientist | Full Stack Developer
Building real-world ML products for business impact.

⭐ If You Found This Project Useful

Give it a star ⭐ and connect to collaborate!