Unstructured and inconsistent
Every show, series or film submitted cast, crew, producers, directors, episodic fee breakdowns and contract details in different PDF layouts.
2021
- Data Engineering
- Automation
- Python
- SQL
- Tableau
Automating revenue and production data extraction for Apple TV
Associate, Analytics & Insights, Cognizant
Built in 2021: dynamic text parsing before the generative AI boomBefore AI tools made unstructured data easy, this took custom engineering. I built the pipeline that turned inconsistent production PDFs into live leadership dashboards.
At a glance
Executive summary
In 2021 I built an end-to-end automated data pipeline for Apple TV.
It converted multi-format PDF production reports into validated data and real-time executive dashboards.
It eliminated manual data entry and streamlined revenue operations.
Where we started
The challenge
Manual bottlenecks
Revenue and operations teams reviewed every PDF by hand and typed metrics into separate database files and dashboards.
Scalability risk
As the original content library grew, manual entry caused reporting delays and raised the risk of errors in key financial calculations.
What I built
The solution
- 01
Diverse PDF production files
- 02
Dynamic Python extraction engine
- 03
Automated SQL validation and staging
- 04
Live Tableau executive dashboard
Inside the pipeline
How it worked
Dynamic PDF extraction engine
I built a Python model that parses PDFs with varying schemas, column structures and text ordering. It extracts cast, crew, production credits, episode counts and financial figures.
The principles
Key design decisions
- Validate before load, never after: bad numbers never reach the dashboard.
- Build for layout variety, not one template: the parser adapts to schema differences.
- Make anomalies visible: exceptions are flagged for a human instead of being silently fixed.
What changed
Impact
End-to-end automation
Manual data entry replaced by ingestion-to-visualization.
Faster decisions
Reporting turnaround cut from days to near real-time on upload.
Higher accuracy
Automated validation checks supported clean financial reporting.
Verified project metric — to be added
PDFs processed or hours saved per month
The tools
Tech stack
- Python
- SQL
- Tableau
- Advanced Excel
Looking back
What I learned
- I learned that validation matters more than extraction.
- Leadership trusts dashboards only when the numbers are provably clean.
- Good automation removes the work people dislike and keeps humans on exceptions.