I get asked fairly often, usually by science PhD students, about transitioning from academia to data science. Here’s a thread of the things I say most often.

Caveats: I finished my PhD. Also it’s from Harvard. I don’t know how many doors opened more bc of the fancy school than anything else, but I assume it’s some % of them.

Also: white, cis, not obviously queer or disabled. YMMV.

I spent ~3 yrs of my PhD casually preparing.

1. Use your projects to build experience (remember you’ll talk about them in interviews). Try to add ONE new DS skill each time. Examples:
- git
- basics of bash scripting / command line
- function-writing best practices
- pandas and numpy
- drake (reproducible workflows in R)

2. Python is more common for DS jobs, but plenty of R jobs exist too. If you know R, you can learn Python.

I used MIT 6.001x (free!) on edX to teach myself Python one summer (comes with bonus computer science basics); there are many other options. https://www.edx.org/course/introduction-to-computer-science-and-programming-7

3. Choosing a course on something valuable (getting/cleaning data, machine learning, git, SQL, linear algebra) and completing it is INFINITELY better than bookmarking endless unread tutorials and trying to figure out the perfect resource to learn from. (This is a self own.)

4. If you can, do an internship. There are internships for grad students, you can probably convince your advisors to let you, and 100% that internship should be paid. A+ way to get a foot in the door before graduating.

Happy to talk about my experience more if helpful.

5. Twitter is great! The DS community here is lovely, especially #rstats folks. I don’t usually do follow-Friday style tweets but I do have a set of people I recommend to folks new to twitter — is this something people already on here would be interested in?

6a. Speaking of Twitter, it’s useful for something I don’t see people talk about often: learning the vibe of the data science community. How people talk, vocabulary, interesting topics, perennial debates, in-jokes. Don’t underestimate the value of ‘sounding’ like an insider.

6b. It’s like learning the unwritten rules of academia, which Twitter is ALSO useful for. The contextual knowledge that people don’t know they know, so they usually won’t think to share it. It can help you go from outsider with nose pressed up against the glass, to inside.

7. The thing you knew I was going to say: portfolios. Pick a thing you’re interested in, get some data, clean it, analyze it, visualize, explain the results, put it online.

More detail: this reddit thread and this podcast.

https://amp.reddit.com/r/datascience/comments/gf9hrs/what_makes_a_good_personal_project_from_the/ https://twitter.com/robinson_es/status/1319366753898692609

8. Insight fellowships: I’ve heard mixed reviews. Can be helpful if you’re not getting bites.

They teach you how to sound like a data scientist & polish your academic experience to make it relevant for data science positions. Also, you get a portfolio project out of it.

9. Avoid Data Camp for learning Did you know they recently reinstated this CEO with basically no changes or consequences? (cw: sexual assault)

https://www.buzzfeednews.com/article/daveyalba/datacamp-sexual-harassment-metoo-tech-startup

Instead, I recommend Coursera or edX!

10. That concludes my most-repeated pieces of advice when people ask me! I hope it’s useful to more people than just the ones who have sent me emails or DMs. I’m happy to expand on any of this, or answer other questions.