Shorter posts, some published and some still being written.
Data people do their best work close to the teams that create the data. Here's what I learned moving them closer to the product.
Five moments, and how each one turns a point of governance into a yes.
Strict residency rules and the biggest AI models pull against each other. For most product AI, the answer is a smaller model that runs where your data lives.
An always-on assistant on a machine at home, split across four repositories with three levels of exposure, and why each piece is shaped that way.
Setting up an LLM-maintained wiki takes an afternoon. Running one for a month is a different story. What drifted, and what I did about it.
Most teams sit lower on the maturity scale than they'd guess, and most of the climb is a decision rather than a rebuild.
Moving from multi-day batch runs to a four-hour pipeline, and from “good enough” to twelve-nines accuracy.