Data specialist bridging finance and engineering — time-series models and finance ML, plus the platform underneath. Most people do one side. I run both.
The shape of a regulated batch build: clinical and claims data through a medallion architecture — raw ingest to governed gold marts, orchestrated in Airflow, transformed with dbt over Iceberg. Auditable at every hop, because a regulator will ask.
The shape of a real-time build: change-data-capture out of the transaction database into Kafka, Flink scoring events in-stream, flags landing in the analytical store seconds after the swipe. Fraud caught while it's still happening.
The specialist core: constructing and maintaining a fixed-income index — market data ingestion, curve and time-series modelling, rebalancing logic — built by someone who came up through finance and speaks the domain, not just the math.
Proof of work: a data platform pointed at the hiring market. Fifteen thousand job boards discovered, scraped, deduplicated and searched — the whole stack above, running in anger.
Deep, structured access to SEC filings for AI agents: XBRL fact-trees, cross-filing SQL, macro data and offline semantic search. Thirty-one tools, public on GitHub.
View on GitHub →All three orchestrators used in anger — comparison opinions on request. The whole thing runs economically on bare metal, not a hyperscaler bill.
I came up through finance.
Then I went to the engine room.
Most can model or run the platform.
I do both — and I ship.
