Short essays on industrial AI, control systems, and the engineering practice that holds them together. Written by Juliet Uzoeshi, for working engineers.
A working engineer's view on where machine intelligence belongs in a plant — and where it absolutely does not. From the PID at the bottom of the stack to the recommendation engine sitting above it, this is the layered architecture I use to keep AI out of trouble and earning its keep.
Read the essayA working engineer's view on where machine intelligence earns its keep in a plant, and the three places it absolutely does not belong.
Most loop migrations fail in the same five places. A pragmatic risk register from someone who has run more S5-to-S7 jobs than is reasonable.
Why so much industrial ML work is really linguistics in disguise, and how thinking about historian data as a corpus changes the modelling choices you make.
If your operator can't argue with the model, the model is not going to last on the line. Notes on what explainability actually has to deliver in industrial settings.
If you'd like a note when something new goes up — or want to talk about a specific topic — drop me a line.