When picking out a topic to learn, I think it makes a ton of sense to learn adjacent topics to what we already know, i.e. things that have moderate degree of overlap to our existing knowledge base.

I think a moderate degree of overlap is necessary. If there's too much overlap, there's little expansion of the mind that happens when we learn a new thing. If there's too little overlap, there's too big of an entry barrier into learning that new thing. Having a moderate degree of overlap helps.

Eugene Yan writes extensively on the topic of data science careers, and I particularly enjoyed the essay he wrote titled Why Have a Data Science Portfolio and What It Shows.

A tl;dr summary of what he has in there:

• The "why" is more than "getting a job". The process of building the portfolio matters more.
• The process involves developing qualities: persistence, continual learning, altruism (to help others).
• The "what it shows" includes both technical skills and those aforementioned qualities.

And a notable quote:

IMHO, traits and skills are a prerequisite to building a great portfolio. And they reinforce each other.

Also:

A portfolio is just an artifact of our skills, traits, and working process. It’s the destination; it’ll take care of itself if we focus on the journey.

Reflects very much the story behind The Score Takes Care Of Itself, by the legendary NFL coach Bill Walsh.

Richard Feynman was a legendary teacher. The Feynman Method is one of his lasting legacies.

Thanks to Farnam Street's blog, we have a very succinct summary:

1. Choose a concept you want to learn about
2. Pretend you are teaching it to a student in grade 6
3. Identify gaps in your explanation; Go back to the source material, to better understand it.
4. Review and simplify (optional)

On that first step, I happen to believe that learning topics adjacent to what you already know is better than learning arbitrary topics (see: Learn adjacent topics).

What do we mean by "thinking mechanistically"? This refers to being able to think mechanistically through data generating processes. In research data science, this likely requires a deep knowledge of the field that one is applying quantitative methods to, i.e. domain expertise. Though deep domain knowledge is usually correlated with doctoral training or many prior years of work experience, a newcomer can compensate for domain knowledge deficits by demonstrating the skill of being able to learn domain knowledge really quickly, or by having complementary breadth of modelling knowledge that has been honed in a diverse set of settings.

For someone who has not yet had the domain expertise to think mechanistically about a problem, they need to Learn how to learn fast.

This was inspired by my participation in the TAO Data Science Panel.

I'm starting to see a bifurcation in research vs business data science.

• Research vs Business Data Science

How this translates to training needs and hiring

• Researchers think mechanistically about the world
• Learn how to learn fast
• Operate outside your pay grade
• Data scientists should know the data generating process behind their data
• Data scientists should learn how to write good code
• Avoiding technical debt with ML pipelines

And notes for managing data scientists:

• How to motivate your problem solvers