Transfer Learning, from Maths to ML

I always thought I would be a mathematician. All of my friends, who were mathematician apprentices with me in grad school, are mathematicians, whether researcher or teacher. The small break I took after my first post-doc to learn new things is still going on ten years later.

When I transitioned to the (data science) industry, it wasn’t clear what was the expected background for a data scientist. No schools or university were proposing classes on the topic. Scientists that were naturally trained in data science, from Astophysics to Bioinformatics or Econometrics, made for the best candidate. This appetence somehow extended to me and my PhD in Pure Mathematics. Ten year laters, with multitude of courses on the topic and clear expectation for the data scientist role, it’s not clear how much an exotic Math PhD brings to the table.

That’s probably why I always take on every chance to give back a bit to researchers trying to make the jump. Lately I was asked to talk to PhD candidates about that transition at Paris V University. Here is the presentation I gave: