Python Data Analysis Utilities — AI-Augmented Data Analysis Portfolio

What This Does

A production-grade Python toolkit for end-to-end data analysis — from messy CSV ingestion to actionable business insights. Built as Week 2 of a 12-week AI-augmented data analysis curriculum.

Business value demonstrated: Analyzed 100K+ orders from the Brazilian Olist e-commerce platform and found that late-delivery rates more than double during the Nov–Mar peak season (8% → 21%), and that a small, identifiable subset of sellers — 47% of the lowest performers — fall below a 70% on-time delivery floor. Both findings translate directly into concrete operational fixes (seasonal capacity planning, a seller scorecard system).

Key Findings

Full analysis with implications and recommendations: notebooks/week2_eda.ipynb

• Late deliveries spike sharply during peak season, climbing from ~14% (Nov 2017) to ~21% (Mar 2018)
• Order growth plateaued after Nov 2017 and hasn't broken its prior peak since
• 47% of the worst-performing sellers (20+ orders) fall below a 70% on-time delivery floor
• Even top-ranked sellers average 21–31 days per delivery — well above the platform median of 10 days
• Order activity concentrates between 10am–9pm, peaking 11am–4pm; overnight hours are largely idle

Portfolio Deliverables

Deliverable	Description	Link
EDA Notebook	Full analysis — 5 visualizations (8 chart panels) and 5 business findings	notebooks/week2_eda.ipynb
Data Utilities Library	36 typed, documented Python functions across 4 modules	src/
Cleaning Pipeline	Auto-encoding detection, null audit, memory optimization	src/data_cleaner.py
EDA Pipeline	Method-chained transformation pipeline	src/eda_pipeline.py
Analysis Functions	GroupBy and time-series analysis functions	src/data_analyzer.py
Loading Layer	Robust multi-format CSV ingestion	src/data_loader.py
Test Suite	Unit tests covering cleaning and analysis modules	tests/

Quick Start (macOS)

git clone https://github.com/AHTISHAM327/python-data-utils.git
cd python-data-utils
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
# Download Olist dataset to data/raw/ (see notebooks/week2_eda.ipynb Section 1)
jupyter notebook notebooks/week2_eda.ipynb

Repository Structure

python-data-utils/
├── src/                          # All analysis modules
│   ├── data_loader.py            # Robust CSV loading with encoding detection (11 functions)
│   ├── data_cleaner.py           # Null audit, datetime conversion, memory optimization (7 functions)
│   ├── data_analyzer.py          # GroupBy + time-series analysis functions (14 functions)
│   ├── eda_pipeline.py           # Method-chained end-to-end pipeline (4 functions)
│   └── reporter.py               # Formatted report generation
├── notebooks/
│   └── week2_eda.ipynb           # Portfolio EDA notebook (100K+ rows, 5 visualizations)
├── tests/                        # Unit tests for all modules
│   ├── test_all.py
│   ├── test_analyzer.py
│   └── test_cleaning_pipeline.py
└── data/                         # raw/ and processed/ (not committed — see .gitignore)

Professional Standards

All source code in src/ follows:

• Type hints on every function (mypy src/ → 0 errors)
• Google-style docstrings (Args, Returns, Raises, Example)
• Structured logging (logging module, no print() statements)
• Error handling (specific exceptions, never bare except:)
• PEP 8 compliance (black + flake8 → 0 warnings)

Key Technical Skills Demonstrated

• Pandas method chaining (.assign(), .pipe(), .query())
• Multi-level GroupBy with named aggregation
• Time-series resampling (.resample("ME")) and rolling windows
• Memory optimization via category dtype conversion
• Automatic encoding detection
• Professional module architecture (src/ package with __init__.py)

About

Built as part of a 12-week AI-augmented data analysis curriculum targeting Upwork freelance clients at $50–$120/hr. Week 3 adds SQL (SQLite + SQLAlchemy). Week 6 adds OpenAI API integration for natural-language insight generation.