Posts

Mantra Make it work, make it nice, make it fast is a nice way of thinking about software development. It’s not perfect because these are not clear cut activities, but it’s still a good mental model of how one should sort their priorities as they approach some programming task. It’s also mostly self explanatory. First, get your basic version going. Then refactor it, improve readability, follow some architeture pattern, etc. Finally, optimize it. ...

Something’s not right A stakeholder pings you on Slack: “The model’s wrong.” You brush it off initially because metrics looked fine, probably just bad luck. Then you check, and it’s not bad luck. It’s bad performance. So you do what teams generally do when faced with an underperforming machine learning system: you rush to try and fix the model itself. Your first reflex to tweak the model In this scenario the model can be anything. A linear regression, an image classifier, an LLM accessed via some third party API or whatever takes some data in and spits out some data out via the interaction of learned parameters. ...

I’m often asked for a way to deal with the barrage of new models/capabilities/evals/buzzwords in AI. This post answers that. Keep in mind that my focus is on stuff that has a relevant and immediate application to industry, which means that there should be a model I can run, an API I can call or something of the sort. I’m not personally interested in staying up to date with academic developments; if you need to improve your signal/noise ratio in arXiv this post will not be super helpful. ...

My educational and professional background is somewhat nonstandard for my work field. I hold a bachelor’s and a master’s degree in Economics, and I spent 8 years working in public policy at CERES, where I led global and regional macroeconomic analysis. I was also responsible for designing and maintaining most of the econometric models. My master’s thesis was a DSGE model for the Uruguayan economy, which ended up being a lot more complicated for me than I like to admit. ...