Dan Brickey
Data Vault 2.0 · Kimball · AI-leveraged data engineering
I build governed, auditable data platforms — the kind teams can trust and auditors can follow. My work pairs disciplined modeling (Data Vault 2.0, Kimball) with modern, AI-leveraged engineering, and I care as much about lineage and documentation as I do about throughput.
Write-ups from the build log: what I predicted, what the numbers came back as, what it cost, and what each run does not prove. Drawn from the specs, build reports, and decision logs the projects were built with, so every claim here has a record behind it.
Ten benchmark submissions re-scored in a shuffled blind pool under a matched effort protocol. Nine of ten moved, all but one upward, and the error tracked how bad each submission looked rather than scattering randomly. A phantom noise floor turned out to be zero, one published finding died, and a headline result shrank by two thirds.
A written pattern library stripped in stages and scored against a frozen spec. Its existence is worth 11 points; its content and phrasing are worth zero, and 592 words did exactly what 709 did. With no library at all, judgment held at 44 of 44 while conventions fell — and at a lower model tier the failure flips to the half that is invisible in review.
A DuckDB platform moved to Snowflake to
measure how much of a migration is
judgment and how much is labor. Zero
files under dbt/models/
changed, four dialect gaps surfaced and
none of them altered output, and content
parity came back clean across 180,533
mart rows. The expensive part was review
bandwidth, not translation.
Claims and clinical data usually live trapped in fragmented, un-auditable legacy systems. This is a governed Data Vault 2.0 + Kimball platform on synthetic claims and clinical data — spec-driven and AI-leveraged end to end, with full historization and lineage you can trace from source to dashboard.
Comparing where to live means wrestling a dozen disagreeing public datasets with no common ground. Gazetteer unifies them into one Data Vault 2.0 platform — cost of living, income, jobs, and health, lined up for the same place and period — so “where's my happy place” gets a real, data-backed answer.