AutoML Platform

Taking Pandas Profiling a step further with heavy inspiration from Autosklearn this is a platform that streamlines ypur ml workflow: from data upload to model deployment.

Features

Data Upload & Exploration

• Drag and drop or browse to upload your datasets
• Get instant insights into your data structure
• Identify and handle missing data automatically
• Smart detection of numerical, categorical, and datetime features
• Visualize your data with various plots

Training and prediction

• Trains multiple ML models
  • Linear Regression
  • Logistic Regression
  • Random Forest
  • Gradient Boosting (XGBoost, LightGBM)
  • Support Vector Machines
  • Neural Networks
• Automated hyperparameter optimization
• Robust model evaluation with k-fold CV
• Side-by-side performance metrics

Architecture

┌───────────────────────────────────────────┐
│       Frontend (streamlit ui)             │
└───────────────────────────────────────────┘
                      ↓
┌──────────────────────────────────────────┐
│              AutoML engine               │
│  performs: EDA                           │
│            Feature Detection             │
│            Preprocessing builder         │
│            Model Selection               │
│            Train + CV                    │
│            SHAP (explainability)         │
└──────────────────────────────────────────┘
                      ↓
┌──────────────────────────────────────────┐
│              Artifacts (model.pkl)       │
└──────────────────────────────────────────┘
                       ↓
┌───────────────────────────────────────────┐
│   fastAPI Backend (prediction endpoint)   │
└───────────────────────────────────────────┘

Installation

Clone the Repository

git clone https://github.com/sonuj04/Auto_ml_platform.git
cd Auto_ml_platform

Create Virtual Environment

python -m venv venv
#on Linux:
source venv/bin/activate

Install Dependencies

pip install -r requirements.txt

Usage

Running the Streamlit App

Run both in separate terminals.

python -m streamlit run frontend/streamlit_app.py

Open your browser to http://localhost:8501

Start the API server:

uvicorn app.main:app --reload

The API will be available at http://localhost:8000