RAG Works Better When Your Data Is Already Tidy
Every retrieval project I've watched go badly went badly for the same reason.
Read the articleBlog
Short notes on data work.
Things I've had to work out, written down before I forget them. Pipelines, platforms, and the parts of AI that turn out to be data engineering. Aimed at senior engineers who would rather skip the intro.
Dagster Pipelines Need Clear Ownership
A job failed at four in the morning. Whose was it?
Read the articleBronze to Silver Is Where Trust Starts
Gold layers get the demos. Bronze to silver is where the numbers get earned.
Read the articleIceberg and Nessie Make Data Changes Safer
Branches for tables, and what that does to the way releases feel.
Read the articlePython Logging That Actually Helps
What to write down, and who you're writing it for.
Read the articleData Engineering Best Practices
Five things I keep coming back to. None of them are clever.
Read the articleWhat I write about
Roughly. It depends on what breaks.
Reliable Pipelines
Testing, failure modes, and what to do at four in the morning.
Platform Thinking
Internal tooling, developer experience, and why year two is harder than year one.
AI in Production
Making it useful without three new services in the stack.