Learning Data Science From the Perspective of a Proficient Developer

7 min readDec 8, 2020

As you know, data science, and more specifically machine learning, is very much en vogue now, so guess what? I decided to enroll in a MOOC to become fluent in data science. But when you start with a strong development background, it’s quite different than when it’s one of your primary experiences with programming. Here’s why.

Dreyfus Model of Skill Acquisition

The Dreyfus model states that learners go through different phases in defining how they perceive a situation, which elements they focus on, how involved they are in the outcome, and how they address the issue. For the purpose of simplicity, I will focus on three faces a developer can exhibit according to the model: the beginner (aka novice), the proficient (aka competent), and the expert (aka master).

Beginners apply recipes they know, regardless of context (no situational awareness). They seek out straightforward rules in order to experiment in a safe environment. Beginners often focus on the easiest tasks, abandoning harder ones.
Proficients can choose among recipes, depending on the context. They focus on experimenting with the rules in different backgrounds, necessitating an environment that provides immediate feedback. They might prefer to avoid using third-party tools more often than it is worth it.
Experts create new recipes for needs, knowing everything depends on the context. They intuitively grasp the situation and what is possible or hard. They often start with the hardest tasks or the one with the worst possible outcome, to mitigate risk.

In nutshell: A different stage means different needs. Although programming can be viewed as a tool for data science and not as data science per se, it can be what occupies you the most in your daily practice (except if you are a theoretician). As a consequence, learning data science as an alleged beginner when you have mastered the practical art of programming will not be that efficient. You would better off focusing on things that are not forthcoming.

Python Is King

As it appears that all data scientists around the world are using Python, you might believe you first need to learn Python programming. There is nothing less true. You don’t actually need to be a Python expert. Just simply know the basics, and you will already almost get it by analogy with other languages: import a module, print some formatted data, create functions if you’d like to follow the DRY principle, deal with loops or branches. You will rarely have to deal with object-oriented programming or design patterns.

If you are a proficient developer and want to master data science, learning Python in depth first is probably a waste of time.

If Python is King, the royal court is actually what I would recommend focusing on, i.e., the ecosystem for data science, and more specifically:

Learning Data Science From the Perspective of a Proficient Developer

Dreyfus Model of Skill Acquisition

Python Is King

Written by Marquisrs