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.

I had some interest in programming. I started automating a few things in R. I wrote a script to build a time series of laws and decrees issued by the Uruguayan government that I never ran again. I also tried to figure out whether weekends hade worse weather (rainier, cloudier, colder) than weekdays, but I failed because the API I was using returned nonsense.

Eventually I started building econuy. It began as a set of functions to automate downloads of economic data, and later grew into a website built with Flask+Bootstrap and hosted on Heroku (very 2016–18 coded). At first I had no real motivation beyond practice.

econuy forced me to learn much of what I now know about programming and building web apps. Just to list a few unrelated things:

  • git
  • OOP
  • buying a domain
  • packaging a library and publishing it on PyPI
  • documentation
  • Github actions
  • SSL

Every new feature I wanted to add forced me to learn something new, and I became more and more interested, enough that I started looking for a career change where I could do more of this.

Conveniently, publishing the econuy website positioned me as someone who a) could program and b) “knew about data,” without having studied for it, which was key to exploring other job opportunities. I landed a few interviews and was able to present the project in some venues.

Meanwhile I started blending econometrics with programming, which naturally led me toward machine learning. I devoured the scikit-learn documentation, in my view one of the best resources for learning classical ML, both the theory and its practical implementation.

I received an offer from CPA Ferrere to lead their data analytics consulting area and I didn’t hesitate. I built the team, brought in clients, and closed projects. I deployed code on client systems. I replaced an app that used IBM Watson for tweet sentiment analysis with a library that ran a transformer under the hood. We even fine-tuned GPT-2 on free Colab to emulate the hate speech directed at Uruguayan female politicians.

At some point I saw Mercado Libre was hiring a Technical Leader in data science. I messaged the person who posted the opportunity on LinkedIn, and I passed all the interviews.

The point of this long story is simply to show how I moved from one role to another that was quite different, and which presumably required a kind of training I didn’t have. For me, the enabler was deciding to create a project that let me both learn and demonstrate that I could do things my 9–5 job never gave me the chance to do.

I believe this is far more valuable than taking a few courses or certifications without much practical application, both in terms of learning and in terms of tangibility. It’s one thing posting on LinkedIn that you completed “Data Science Essentials” on some course platform, which people can have a hard time knowing whether it’s just smoke and mirrors; it’s another thing entirely to show something you’ve built that others can actually use.