EngAhmed5/Autonomous-Data-Science-System

An advanced agentic ML System automating the end-to-end data science lifecycle. Integrates LLM reasoning with AutoML to deliver production-ready models and diagnostic PDF reports via automated Gmail delivery. Final project for Orange Digital Center (ODC) Egypt (AI Agents & LLMs track). # Autonomous Data Science System (ADSS) > **Data Science on Autopilot: An end-to-end agentic ML orchestrator designed to transform raw data into production-ready models and executive-level PDF reports.** --- ## The Team This project was collaboratively developed as a **Team Project** for the **Orange Digital Center (ODC) Egypt** (AI Agents & LLMs Track) by: * **Ahmed Ismail El Sayed** * **Ahmed Mohamed Hussein** --- ## Project Overview The **Autonomous Data Science System (ADSS)** is a sophisticated multi-agent framework that automates the entire machine learning lifecycle. By combining the reasoning capabilities of **Large Language Models (LLMs)** with the computational power of **Scikit-Learn**, ADSS handles everything from data ingestion to sending a professional diagnostic report directly to your inbox. ### Agentic Workflow The system employs an **Orchestrator-Agent Architecture** consisting of specialized autonomous units: * ** EDA Agent:** Generates statistical summaries, detects outliers, and creates a comprehensive visualization suite. * ** Preprocessing Agent:** Dynamically engineers features, handles scaling, and encodes variables based on real-time data inspection. * ** ML Tuning Agent:** Optimizes model performance via `RandomizedSearchCV`, comparing multiple algorithms to find the best fit. * ** Error Analysis Agent:** Interprets complex technical metrics into high-level, human-readable diagnostic insights using LLMs. * ** Documentation Agent:** Programmatically builds professional multi-page PDF reports using `ReportLab`. * ** Email Agent:** Automates the final delivery of reports and artifacts via Gmail. --- ## Project Demos ### Streamlit Pipeline Interface github.com --- ## Repository Structure ```bash │ requirements.txt │ ├───Data # Sample datasets used for testing │ Bank Customer Churn Prediction.csv │ Car details …