Back to portfolio

Automated Analytics

CognitoEDA - Automated Exploratory Data Analysis Platform

Developed CognitoEDA, an AI-powered exploratory data analysis platform using Python and Streamlit for automated dataset intelligence, visualization, and interactive analytics on CSV datasets.

Overview

Challenge

Preliminary exploratory data analysis is repetitive and time-consuming, while unexpected dataset schema discrepancies often crash downstream machine learning pipelines.

Solution

Built an automated dataset intelligence platform that profiles tabular datasets, generates health scores, and flags training impediments.

Status

Analytics Platform

Architecture

CLICK ANY STAGE TO INSPECT ITS INPUT / OUTPUT SPECDATA FLOW 5 STAGES ✓

Implementation

  • Coded an automated profiling engine that evaluates data distributions and missing value ratios in a single pass.
  • Designed a data health scoring algorithm that rates dataset training readiness based on class imbalances.
  • Built interactive correlation maps and statistical summaries into a Streamlit dashboard.

Results

  • Automated initial dataset exploratory analysis routines into a single click.
  • Eliminated pipeline crashes by detecting schema mismatches and missing values before model training.
  • Decreased data preprocessing troubleshooting overhead by flagging distribution anomalies.

Lessons Learned

  • Automation should prioritize flagging data errors over generating excessive generic charts.
  • A numeric health score drastically speeds up the go/no-go training decision for developers.
  • Strict input validation prevents malformed files from corrupting statistical aggregation metrics.

Tech Stack

PythonStreamlitPandasScikit-learnStatistical Profiling