./about — who keeps this notebook

RobotNurse is one working engineer, on-site, with a toolkit.

robotnurse.io is run by RobotNurse, an engineer who builds and runs local AI systems for a living. The service is simple: I come to your home or shop and set up powerful language models on hardware you own — so your data stays on your machine, and the machine stays yours.

[ the service ]

The work is hands-on local AI setup for hobbyists and small shops. That means installing, tuning, and handing over open-weight models that run entirely on your own hardware — private, offline, no subscription to a cloud API. You get a working system, a walkthrough, and a plain-language guide for keeping it healthy after I leave.

I work on the machines people actually buy: Mac Studio and Mac mini boxes (unified memory is a gift for local models), GPU PCs (Nvidia and AMD builds), and compact Spark-class AI devices and mini-PCs. If you already own hardware, I tune what you have. If you don't, I'll tell you honestly what's worth buying — the buyers' notes are the same advice I give on-site.

[ how i work ]

1. 01

   Fixed scope, no retainer required

   Every job is scoped before I arrive. You know what gets done; there is no open-ended hourly meter running in your living room.

2. 02

   Your hardware, your weights

   Nothing is rented back to you. When the job is done, the system, the models, and the config files are all yours.

3. 03

   Tested twice, claimed once

   Every setup is verified working before handover, and every claim on this site has a log file behind it — not a marketing brief.

[ the notebook ]

robotnurse.io exists because most writing about running LLMs either oversells it or arrives pre-edited by a marketing department. This is neither. It's a notebook, kept mostly in public, about the slow, unglamorous work of getting open-weight models to run well on hardware that belongs to you. The name is a joke about wandering, fixing, and staying on shift. Nothing here is sponsored, nothing is promoted, and nothing appears because a vendor wanted it to. Every benchmark is reproducible; every claim has a log file behind it. If you read top to bottom you'll find more dead ends than victories, because that's what honest field notes look like.

[ get in touch ]