Conversational AI agent for automated exploratory data analysis

focus

Artificial Intelligence

tools

Python

Report

Visualization

Streamlit

code

Github

Conversational AI agent for automated exploratory data analysis, enabling users to upload datasets and receive AI-generated insights, visualizations, and follow-up analysis through natural language interaction powered by Claude API.

PURPOSE

Automate exploratory data analysis by enabling users to upload datasets and receive AI-generated insights, visualizations, and statistical summaries through natural language interaction, eliminating manual analysis workflows and reducing time-to-insight from hours to seconds.

‍

OBJECTIVE

Enable non-technical users to perform comprehensive data analysis through conversational AI by uploading CSV files, receiving automated statistical summaries with matplotlib visualizations, and asking follow-up questions with context-aware responses tracked across 10-turn conversations with token management and session state persistence.

PROCESS

Implement direct Anthropic Claude API integration with custom tool orchestration for visualization generation, building keyword-based intent detection to distinguish analytical questions from chart requests, optimizing token usage by sending summary statistics rather than raw data, and deploying full-stack Streamlit application with conversation management, error handling, and Railway hosting for production access.

‍

OUTPUT

Deployed production conversational AI agent supporting 4 sample datasets (Iris, Wine, Diabetes, Breast Cancer) plus custom CSV uploads, featuring automated insights with dual matplotlib chart generation, context-aware follow-up responses with token tracking (50K limit), turn-based conversation limits (10 turns), and real-time code execution with session state management. Successfully demonstrates AI API integration fundamentals, prompt engineering, and tool use orchestration for portfolio differentiation.