Lessons Learned from Two Years as a Data Scientist

Intro

I finally escaped from (grad) school in 2019, spent two months interning as an assistant trader at FTX, and have since spent the last two years working as a data scientist, with the bulk of that time at Microsoft and the last two months at Anthropic. Not gonna lie — my time as a data scientist has been pretty awesome. I was technically on a product team at Microsoft in DevDiv working on improving developer tools like VS Code, but in practice it was like a flush research job with tons of freedom (to first approximation my instructions for the first six months were, “Fix bugs using machine learning. Go.”) I got to work with cutting edge transformer models and their application to source code, the most famous of which is autocompletion ala intellisense, tabnine, and Codex. Every two months the amount of compute I have access to has doubled, starting out with a 2013 macbook air and ending with a cluster with thousands of A100s (which, to be fair, I share with more people than my old laptop), and I got to collaborate on ten different papers.

I was extremely green when I started out. I hadn’t used python since cs1 as a college freshman, excluding one month where I worked through fast.ai and leetcode. Maybe the most telling anecdote is that I spent an embarrassingly long time using vim as my exclusive file editor… while not knowing any of the vim shortcuts besides :wq (write and quit).

Read in full here:

https://dawndrain.github.io/braindrain/two_years.html

This thread was posted by one of our members via one of our news source trackers.