PipeHouse
Early accessData insights, without the data team.
Managed Snowflake, dbt, and ingestion, plus an MCP server so an AI agent queries tested models instead of raw tables.
Data infrastructure for AI-ready finance and revenue teams
I spent a decade as the person who had to defend the number in the board deck. Now I build the infrastructure underneath it.
Point a language model at an ungoverned warehouse and it will confidently return a figure nobody can stand behind. The fix is not a bigger model. It is a harness: tested dbt models, one definition per metric, and a query surface that only exposes what has already been validated.
Data Pipeline Builders
A productized data platform for mid-market companies, roughly $10M to $200M in revenue, that need a data team's output without the headcount to match. Three pieces, one thesis.
PipeQL
BetaCollaborative SQL for your team.
A shared SQL editor for teams that need version history, saved queries, and a home after PopSQL shut down.
Try it →PipeAcademy
In developmentData engineering, taught by practitioners.
Courses and training on dbt, Snowflake, and the modern data stack, built from production work rather than slideware.
The operator and the engineer
I have built the board deck
FP&A and go-to-market analytics at LinkedIn, Modern Health, and beyond. I know which numbers a CFO gets asked about, and which ones quietly break the week before the board meets.
And the model behind it
Snowflake, dbt, dlt, and Prefect in production. Most data infrastructure founders can do one side of this. The value is in doing both, because the metric definition and the SQL are the same decision.
SnowflakedbtdltPrefectPythonSQLMCP