Custom AI Training Interface for Tabular Data (End-to-End ML Pipeline)
Developed an interactive web-based machine learning training interface that allows users to upload structured CSV datasets, select target labels and feature columns, and train classification models without writing code. The application supports Logistic Regression and Random Forest models, includes automated data preprocessing (handling missing values, one-hot encoding categorical variables, and feature scaling), and performs configurable train–validation splits. The system provides real-time model evaluation through accuracy, weighted F1-score, confusion matrices, and detailed classification reports. Trained models and their preprocessing pipelines can be saved and downloaded as reusable artifacts. A single-row inference playground enables users to input new data and generate predictions using the trained model, ensuring end-to-end consistency between training and deployment. Built using Python, Streamlit, and scikit-learn, this project demonstrates the full machine learning lifecyc