Conversational Data Profiler

Worked on a small but ambitious project for Data Science. The vertical is completely new to what we have worked so far and there are lot of data to analyze to create meaningful insights. The files are very big and the relationship between different data set were not established and there are lot of ambiguity because of many to many cardinality. It was hard to even get a single insight with all data combined.

WebApp powered by a smart AI bot that enables Exploratory Data Analysis (EDA) without requiring Python or traditional BI tools. The application accepts multiple data formats, including Excel, CSV, spreadsheets, and SQL connectors. Upon uploading, the interface displays a data preview and profile, covering rows, columns, and unique values. We integrated Great Expectations to evaluate data quality across five key pillars. A built-in AI chat sidebar allows users to generate EDA reports, identify primary and foreign key candidates, explore dataset relationships, and receive BI tool recommendations. During development, we considered consolidating everything into the AI chat alone. However, architectural analysis revealed that splitting the two components was essential - the WebApp handles large data loading and profiling, which can be resource-intensive and slower, while the AI chat remains responsive and spontaneous, delivering instant answers to user queries without performance compromise