20 Tips For Becoming a Data Scientist

1. Expect the worst. Data will always be dirtier than you expect. Always.
2. Explore before you dive in. There is nothing worse than working hard but moving in the wrong direction. Start with data exploration and descriptive statistics before you start building models.
3. Start with simple models. Start with the simplest approach possible. If that's not enough, then you can get the more powerful tools out.
4. Random forest. You can probably use random forest for that. Yes, that too. And that.
5. Team work. You'll spend a lot of time banging your head against a wall if you go at it alone -- embrace collaboration. You are only as successful as your team.
6. Replace the word "I" with "we". This is a rehash of point 5, but it bears repeating: you are only as successful as your team.
7. Learn software engineering. You need to be able to productionalize your models. Creating models in R isn't enough these days; you need to build models and software if you want your products to scale.
8. Continuous improvement. You won't get it right on your first try. Trust me. Your goal is to be fast, get something done, and then improve it iteratively. Be agile.
9. Get really good at tuning models quickly by hand. Get a deep understanding of your frequently-used models (random forest!) and develop the intuition to tune them quickly by hand. Ain't nobody got time for grid search when you're building models.
10. Get really good at automating your pipelines. This is where you use grid search. And airflow. Or maybe luigi. Whatever the tool, automate the hell out of your pipelines to eliminate doing the same work over and over again.
11. Stop speaking like a statistician. Nobody understands you and it won't make people like you. Just be a normal person. Be cool. Yeah, like the Fonz. Cool.
12. A bigger computer is never the answer. Okay, sometimes it is, but not usually. When your models blow up the server, think about how you can more efficiently handle memory and parallel processes. Consider dimensionality reduction. Break the problem down into smaller problems that can be easily solved and put back together. Don't resort to blaming the hardware -- you'll be out of runway (and money) quickly if that is your only solution.
13. Help others understand data science. Most of it isn't that complicated: gather data, recognize patterns, and predict what will happen next. Explain it to people using plain English, and...
14. Tell stories. No, not campfire stories. Actually, that sounds like fun. If you like telling campfire stories, come work for me. But also tell stories using data. Stories are how humans connect. Use the data to tell a story and you will be able to relate your results to anyone.
15. Don't fall in love with you work. It's probably not that good, and even if it is, nobody cares. You need to fall in love with the results. Put your ego aside and stay results focused.
16. Be client focused. This is an extension of the last point, but it's important to always view your results from the clients' perspective. They don't care that you used the latest deep Bayesian neural Boltzmann long-short tractor-bottle machine. They care about how you help them improve their business and make money. Put yourself in their shoes during every step of the process. p.s. I hope you didn't try to google that.
17. Drink beer with other data scientists. And normal people -- I hear they like beer too. Okay, you don't have to drink (but it is fun), but you do have to network and connect with other people. You'll learn a lot by understanding other people's perspectives. And you might be able to help a few people along their journey of life while you're at it, which is pretty cool too.
18. Learn behavioral science. Neuroscience. Psychology. All that stuff. If you don't work with humans, or are yourself a robot, don't worry about this one. Otherwise, this will be big in the future, and you'll want to learn it if you are going to stay ahead of the curve.
19. Keep learning. You can innovate or you can become obsolete. Read a lot. Books, blogs, textbooks, articles. Anything that will expand your perspective and knowledge base. 95% of you current skills will be absolutely common in 5 years -- what will differentiate you then?
20. Thanks for reading this far. I only had 19 tips. It was a trick all along. Hope you enjoyed. Now share this with your friends and go learn something!

Here are 5 books to checkout if need some help getting started:

1. Scrum - Sutherland
2. The Elements of Statistical Learning - Hastie
3. Evolutionary Psychology - Buss
4. The Power of Habit - Duhigg
5. Growth Hacker Marketing - Holiday