Forty years of my grandfather's handwritten notebooks, and an attempt at transcribing them that works about half the time and is still worth having.
Reading my grandfather's notebooks with a model that mostly fails
Jomar Salcedo
@jomar-salcedo
Builder profile
@jomar-salcedo
Two and a half years into a job doing machine learning at a company where most of the actual work is data plumbing and expectation management, which nobody told me at university. I am fine at models and I am learning that the model is rarely the thing that decides whether something works. Most of my week is spent on evaluation — building the set you measure against, arguing about what counts as correct — which is much less interesting to explain at parties and much more useful. At home I build small things that classify things around my flat, mostly because it is the fastest way to find out what I do not understand. Still quite junior and trying not to pretend otherwise.
Joined July 2026
Forty years of my grandfather's handwritten notebooks, and an attempt at transcribing them that works about half the time and is still worth having.
Jomar Salcedo
@jomar-salcedo

Three cats, one bowl, and one of them is on a prescription diet, so I taught a camera to tell them apart and log who eats from which bowl.
Jomar Salcedo
@jomar-salcedo
A model that decides which team a support ticket should go to, replacing a routing rule set that had grown to several hundred keyword rules.
Jomar Salcedo
@jomar-salcedo