FAQ
How did you get from Physics into ML?
ML is very much an empirical discipline, which means the we mostly learn from lots and lots of experimentation — and not just random experimentation, but hypothesis-driven experimentation, i.e. following the Scientific Method. As a Physicist, this way of thinking and making progress felt quite natural, and I decided to dig deeper.
I took my first ML courses while still a grad student in Helsinki. When I was a postdoc at Argonne, I spent some of my personal time building a simple classification model for astronomical images, and later helped a team at the Argonne Leadership Computing Facility build a deep neural net for brain image segmentation.
Eventually I landed my first industrial job by meeting my future manager at a networking event at the University of Chicago, and became Data Scientist at JP Morgan Chase’s “Digital Intelligence” group. From there I pivoted to risk modeling at Amazon and eventually recommender systems / ranking models at Meta.
If you ever
- see a blurb “Customers like you save an average of $X a month” in your Chase app,
- get a package from Amazon that has a Dangers Goods label on it,
- see an extremely good video recommendation in your Facebook feed,
chances are I had my hands on the ML model behind it.
Why do you write?
I mostly write to learn. This is also known as the Feynman technique. If you can’t write it down in a paragraph, that means that you probably haven’t understood it well enough yet. Besides, writing for an audience online is a great way to build a brand and create visibility for future opportunities. It’s an investment.
Do you have any tips for getting started with writing?
Yes! Check out my post here:
6 Simple Lessons That Will Help You Start Your Writing Side Gig
What’s the architecture of this site?
It’s using Jekyll, Github Pages, and a Google domain. You can find the source code on my github.
What’s the best way to get in touch?
You can find me on LinkedIn.
What are some questions that you’re being asked a frequently?
Link.