Case study · Warehouse · AI · Oil & gas
Land data stack
Reverse-engineering a land system nobody could fully explain — and turning it into reporting views people trust.
Problem
Land data is where energy reporting projects go to stall. Leases, tracts, agreements, and interests reference each other across systems in ways that only make sense if someone documented them — and here, almost nobody had. The client needed a cross-reference system for land entities, but the source system offered limited insight and very little documentation to build from. The relationships had to be discovered before they could be modeled.
Approach
Rather than interviewing my way through every undocumented join, I built Python-based AI lineage tracing to analyze the source data itself — profiling entities, testing candidate relationships, and reverse-engineering how land records actually connected in practice. The discovered lineage became the blueprint for the design.
On top of that foundation I architected and developed the land-entity cross-reference system in Snowflake, carrying it through to the reporting views the business consumes. I stayed on after development to maintain the stack — supporting reliability, reporting accuracy, and ongoing enhancement as the land organization's needs evolved.
Architecture
Outcome
The land organization got a cross-reference system grounded in how its data actually behaves, not how anyone assumed it should. Relationships that once lived in a handful of heads are now encoded in the model itself, and land reporting runs on views with a maintained, documented foundation underneath them.