About Me

I've worked with data since 2013, mostly in places where a wrong number has consequences. At a regional retailer I took a SAP BW warehouse from one country to a multi-country operation — multi-currency, with a security model where each country sees its own numbers and nothing else.

At a bank, I rebuilt the risk data warehouse during a core banking migration: reconstructing the feeds while the source system underneath was being replaced. The part I'm proudest of came out of that migration. The IFRS 9 expected-loss calculation ran on SQL Server — but once we moved to Hadoop, relational code couldn't do the job anymore. We tried Python; on ~200,000 loans it ran out of room. PySpark was what actually solved it. I didn't reach for distributed computing because it was fashionable — I reached for it because nothing else finished the run.

Today I'm a senior data engineer at Credicorp Bank. Outside the bank I build my own product: RanchOS, a herd-management system used by real operations — everything from the data model to the analytics pipeline on BigQuery and dbt. It's where I test in production the decisions I write about here. I'm also working through the DMBOK with a DAMA study group, applying it to a system I actually own, which turns out to be a very different exercise than applying it to a slide.