Tips for aspiring data scientists: start a blog

Ironically, that tweet suggests advice I've given at least three dozen times, but I still have not written a post about that. I have given that advice to almost all aspiring data scientists who asked me what to do to find a job: start a blog and write about data science.

What could you write if you are not yet working as a data scientist? Here are some possible topics (each one attached to the examples of my own blog):

Analysis of data sets that you find interesting

• Intuitive explanations of concepts that you have recently learned
• Exploitations of how to make a specific piece of code faster
• Ads about open source projects that you've released
• Links to the interactive applications that you created
• Sharing a rewrite of conferences or meetings that you have participated
• Express opinions on data science or educational practice

In a future post, I would like to share tips on how to make the data science blog (how to choose a topic, structure an entry and publish with blogdown). But here I will concentrate on the why. If you are in the middle of an employment survey, you are probably very busy (especially if you are currently employed) and the blog is a substantial commitment. So, here I will present three reasons why a data science blog is worth your time.

Practice data analysis and communication about it

If you expect to be a data scientist, you (presumably) are not yet one. A blog is your opportunity to practice the relevant skills.

Data cleansing: One of the advantages of working with a variety of data sets is that you learn to collect data "as it is", either in the form of a supplementary file of a newspaper article or a movie script.

Statistics: Working with unknown data allows you to put statistical methods into practice, and writing posts that communicate and teach concepts helps build your own understanding

Machine learning: there is a big difference between using a forecasting algorithm once and using it in several problems, as well as understanding why to choose one instead of another.

Visualization: Having an audience for your graphics encourages you to start polishing and building your personal style.

Communication: You gain experience writing and get practice structuring an argument based on data. This is probably the most relevant skill that blogs develop, since it is difficult to practice in other places, and is an essential part of any career in data science.

I can not emphasize how important this kind of practice is. No matter how many Coursera, DataCamp or bootcamp courses, you still need to have experience in applying these tools to real problems. This is not exclusive to data science: be what you do professionally today, I'm sure you're better at that now than when you finished your classes.

One of the great thrills of a data science blog is that, unlike a course, competition or job, you can analyze any data set you want! Nobody was going to pay me to analyze the plot of Love Actually or the titles of Hacker News. Whatever your fun or interest, you can find relevant data and write some posts about it.

Create a portfolio of work and skills

Graphic designers are usually not evaluated on the basis of points in the curriculum or in statements in a job interview: they share a portfolio with examples of their work. I think the field of data science is changing in the same direction: the easiest way to evaluate a candidate is to see some examples of data analysis that they performed.

The blog is a good fit to show your skills because, unlike a technical interview, you can put your best foot forward. What other skills do you take pride in?

If you have the ability to see data, perform some analysis with some attractive and informative graphics ("See an interactive view of mushroom populations in the United States")

If you are good at teaching and communicating, write some lessons on statistical concepts ("See an intuitive explanation of the PCA")

If you have a special talent for machine learning models, blog about some predictive realizations ("I was able to determine a dog's breed from a photo with 95% accuracy")

If you are an experienced programmer, announce open source projects that you have developed and share examples of how they can be used ("With my sparks package, you can load CSV datasets into the Spark 10X faster than previous methods")

If your real expertise is in a specific domain, try conc go into it ("See how penguin stocks do not need to be perfect, in general, when I'm evaluating a candidate, encourage me to see what he shares publicly, even if you expect employers to watch your work, That does not mean that he needs to be perfect, that it is not polished or finished, and sharing everything is almost always better than sharing nothing. 

In this post, I shared how I got my current job, when a Stack Overflow engineer saw one of my posts and I was looking for it, which certainly qualifies as an accident but the more public jobs you do, the greater the chance of a strange accident like that: from someone perceiving your work and aiming at a job opportunity, or from someone who is interviewing you have heard Talk about the work you did ... Blogging is not just to advertise to employers, you can also create a network of colleagues and data scientists colleagues, which helps both to find r a job as in your future career. (I think #rstats users on Twitter are an especially incredible community). 

A good example of someone who succeeded in this strategy is my colleague Julia Silge, who started her excellent blog while looking to change her career to data science, and both got a job and built productive relationships through it. Get feedback and evaluation. Currently, we are looking for your first job as a data scientist. You completed all relevant DataCamp courses, worked on some books and practiced some analysis. But you still do not feel ready, or maybe your applications and interviews have not been worth it, and you decide that you need a little more practice.

What should you do next? What skills could you improve? It is difficult to tell when you are developing a new set of skills to where you are and what you should learn right away. This is one of the challenges of autonomous learning, instead of working with a teacher or a mentor. A blog is a way to get this kind of feedback from others in the field. This may sound scary, as if it could receive an avalanche of criticism that will take you away from a topic. But in practice, in general, you may realize that you are not ready before finishing a blog post.

For example, even if you are familiar with the fundamentals of random forests, you may find that you do not it is possible to achieve precision. you expected in a Kaggle data set - and you have the opportunity to postpone your blog post until you have learned more. The important thing is commitment: it's easy to think "I could probably write this if I wanted to," but it's harder to try to write it. Which of your skills are more developed, or more important, than you thought it was? This is the positive side of self-evaluation. After sharing some analyzes and codes, you will probably discover that you were underestimating in some areas. This affects everyone, but it is especially important for the doctorate. 

Students, who spend several years becoming experts in a specific subject while surrounded by people who are already specialists - a recipe for imposter syndrome. For example, I took the principles of Bayes' empirical estimation when I was a graduate student and it was a simplification of the "real" Bayesian analysis, I assumed it was not worth talking about. But once I wrote about empirical Bayes, I learned that these posts had a substantial audience and that there is a real lack of intuitive explanations for the subject.

I just expanded the posts in an electronic book: most of the material in the book would never qualify for an academic publication, but it would still be worth sharing with the rest of the world. A question I would like to ask doctoral students and anyone else. With a difficult but close experience, it is "What is the simplest thing that you understand that almost nobody outside of your field does?" This is a recipe for an optimal and useful post on the blog. Conclusion One of the most difficult mental barriers to start a blog is the concern that you are "screaming in a vacuum".

If you have not yet developed an audience, it is possible that almost nobody reads your postures in the blog, so why put the work in them? First of all, many of the advantages I described above are as useful as you have ten followers on Twitter or ten. one thousand. You can still practice your analysis and writing skills and point potential employers for your work. And that helps you acquire the habit of sharing work publicly, which will become increasingly relevant as your network grows.

Second, it is at that point that people who are already members of the science community of data can help. My promise is this: if you are at the beginning of your career as a data scientist and ini a blog related to the data, send a link to me in @drob and I will tweet about your first post (actually, the offer is good for each of its three first places). Do not worry if you press or "good enough to share" - just share the first job you find interesting! 3 I have a decent sized audience and, more importantly, my followers include many data scientists who are very supportive of beginners. and they are interested in promoting their work.