GS
Back to projects
🧹

Data Cleaning & Transformation Pipeline

A reusable Python pipeline that catches data problems before they become reporting problems.

PythonPandasNumPy
View repository

Problem Statement

Downstream reports were breaking or showing wrong totals because raw data had inconsistent types, nulls, and outliers.

Dataset Overview

10,000+ raw records with mixed data types, missing values, and inconsistent formatting from multiple source systems.

Cleaning Process

  • Built type-validation and null-handling functions with Pandas
  • Detected and flagged statistical outliers with NumPy
  • Logged every transformation for auditability

SQL / Analysis Queries

df.groupby('category').agg(['mean','std']) for outlier thresholds
Vectorized NumPy operations replacing row-by-row loops

Business Insights

  • A single upstream export was responsible for most bad records
  • Outlier volume dropped sharply once source validation was added

Results

Cut reporting errors by 95% across downstream dashboards
Pipeline now runs unattended on every new data drop