Case Studies
Engineering Blog
Technical write-ups from real projects. Each post covers a decision I made, why I made it, and what the outcome looked like in production.
How I Unified 5 ERPs into Golden Customer Records with Splink
When five separate ERP systems each have their own version of "Boeing," you need more than a VLOOKUP. Here is how I used probabilistic record linkage to resolve 82 duplicates from 333 source records into 251 golden customer entities.
Re-engineering 50 Notebooks into a 2-Minute DAB Pipeline on Databricks
A financial data startup had a production pipeline built from approximately 50 sequential notebooks running for over 90 minutes. I rebuilt it as a modular DAB project with Lakeflow DLT, cutting runtime to under 2 minutes with zero accounting differences.
Why I Chose dbt Core over dbt Cloud for This Project
dbt Cloud is excellent, but it was not the right fit for a self-hosted data platform on Railway with PostgreSQL. Here is how I evaluated both options, what tipped the decision, and what I would choose differently for a Databricks-native stack.