Practical wins from AI, microcontrollers, sensors and edge-AI chips on the shop floor. Written in our own words, with every source linked and an honest label on how strong the evidence is.
Brian Roemmele reports wiring an AI agent into an existing flexographic carton-box printing line, with small local models watching the machine and cameras feeding it visual data.
Evidence: Self-reported by the builder on X. Early, no independent verification yet. · via @BrianRoemmele
Brian Roemmele showed a Raspberry Pi with an attached camera running the open LLaVA 1.5 model locally to describe and recognize what passed in front of it.
Evidence: Hobbyist demo shared on X (2023). Not a factory deployment. · via @BrianRoemmele
Neo Industries, a small machine shop in Colleyville, Texas, used automatic production monitoring to see how much of each cycle every tool was responsible for, then optimized tool by tool.
Evidence: Vendor-published customer story with named shop owner.
MSP Manufacturing in Bloomington, Indiana, uses an AI CAM plug-in to draft most of each program, with experienced programmers finishing and proving out the first run.
Evidence: Vendor-published Q&A with named company president.
At Bosch's Homburg plant, milling machines were retrofitted with sensors so a model could judge tool wear instead of changing tools on a fixed, conservative schedule.
Evidence: Published pilot project (German Green-AI Hub) at a large manufacturer.
A print shop mounted a small sensor board inside an old Xerox iGen4 to learn its normal vibration and flag anything unusual, plus monitor solvent fumes.
Evidence: Vendor tutorial based on a real print-shop scenario. No downtime results published.
A precision screw-machining company making millimeter-sized metal implants brought outsourced inspection in-house with a multi-camera AI station.
Evidence: Vendor-published case study; client anonymized.
A detailed guide on giving an old CNC with no data output a spindle-load monitor using a clamp-on current sensor and an ESP32.
Evidence: How-to guide from an engineering blog. A design write-up, not a measured deployment.
One short email a week: a tool update, a real-world win (with numbers and sources), and a deal or trial worth knowing about. No hype.