Now Playing: “Running” by Helado Negro
Quick Overview
I joined RVNG Intl. in summer 2025 to build a centralized data foundation for the label’s revenue, sales, and catalog data.
My work focused on data infrastructure: consolidating more than ten years of historical records, standardizing data from different platforms, and building automated pipelines and a relational database for future analysis.
The Data Problem
RVNG’s data was spread across digital distributors, Shopify, Bandcamp, and other operational sources.
Each platform used different file formats, identifiers, currencies, time periods, and reporting conventions. Answering a question about an artist, release, or market often required manually combining several exports.
Before the label could build reliable dashboards or perform deeper analysis, it needed a consistent way to ingest, clean, and connect these records.
My Role
I built Python and SQL pipelines that consolidated more than ten years of digital revenue and direct-to-consumer sales data.
The pipelines standardized fields such as dates, currencies, product identifiers, territories, and platform names before loading the records into Cloud SQL.
I also designed the relational structure connecting transactions with artists, tracks, releases, platforms, and territories.
How I Approached It
I began by studying how each source represented the same underlying concepts.
A product in Shopify did not necessarily have the same identifier or structure as a release in Bandcamp or a recording in a distributor report. Some relationships could be created through stable identifiers, while others required carefully defined mapping rules.
I separated the workflow into ingestion, validation, transformation, and loading stages. This made it easier to identify where a data-quality problem had entered the system and to add new sources without rewriting the entire pipeline.
I also treated the schema as a shared analytical foundation rather than storage for one particular dashboard. The goal was to support future questions that had not yet been defined.
Outcome
The project brought more than ten years of fragmented revenue and sales data into an automated and scalable pipeline.
The resulting Cloud SQL database created a unified source for catalog and market analysis and reduced the need to manually combine historical exports for every new question.
It also established the data foundation that I continued developing during my fall internship.
What I Learned
This was my first time building infrastructure around a long and imperfect business history.
I learned that most data-engineering decisions are also decisions about meaning. Before two datasets can be joined, someone has to determine whether their identifiers, dates, and categories actually describe the same thing.
The experience made me more careful about data lineage, validation, and designing systems for questions that may come later.
Tools
Python, Pandas, SQL, Google Cloud Storage, Cloud SQL, ETL, Data Modeling