So you want to get into data science?

September 08, 2024

2024   ·   data-science     ·   posts  

Images taken from unsplash.com .

Introduction

As with all the content on this website and in my blog posts, the opinions expressed here are my own and not those of my employer. More and more often recently, I’ve been asked by students and others “So, what do I need to do to get a job as a data scientist?”. I’ve realised, although I have lots of thoughts on this, the answers I’ve been giving have been rather rambling and unstructured, with more points than any person can reasonably hold on to while talking to me at a careers fair. So I decided to put these thoughts into this short post to bring these notions together in one place. This article will naturally be biased towards someone with a degree in natural sciences (e.g. physics), who has had some experience in programming and statistical analysis, and also inevitably biased towards my personal opinions and recollections.

Do you really want to get into data science?

As always before embarking on anything new which will take up a lot of your time, you should yourself two start with two key questions:

  1. Why do I want to do this?
  2. Do I really know what I’m getting into?

I highlight these two questions because putting a lot of effort into breaking into data science before realising that it isn’t what you want to do, or it isn’t really what you thought it would be, would be less than ideal. Data science is often not as glamorous as it is depicted, although the perception is now a bit more grounded than it was in the days of “the sexiest job of the 21st century”. As a data scientist, you are required to undertake several different roles at once, stay up to date with a technology landscape that changes extremely quickly, work with requirements that are often unclear and shifting, and, most crucially, you are required to make an impact on the business; to senior management you are a new investment, and they want a return on that investment. Many technology and software-based roles also have these aspects to various degrees, but these are all distinct realities of being a data scientist. Part of this comes from data scientist roles looking different across different organisations and how rapid developments in the world of data and AI has led to high expectations from business leaders. However, my personal opinion is that if you want a job that uses programming and statistical modelling to solve problems, and you want to convince business leaders you’ve solved those problems, you could do far worse than a role as a data scientist.

Roles in data beyond data science

A point sometimes neglected is that the volume and variety of data roles beyond that of data scientist are growing year-on-year. As data science practices mature in businesses, the range of professions is increasing, with roles that are more defined than before. This trend seems to closely follow business consensus that you need a much more sophisticated data strategy to really transform a business than just training a model. Some roles, especially data engineers and machine learning engineers, are more sought after than data scientists, and will commonly pay better. There are also roles that require a different range of skills and still provide substantial business value. For example, data analysts can give great insights from data and clearly inform business decision-making, without the need for python and complex algorithms. Additionally, it’s becoming more commonplace for individuals to pivot from related data roles into data science. In short, it will be well worth you’re time exploring which one of these roles you think might best fit your skill set and what you enjoy doing.

How would I go about learning data science again

So, you’ve looked the other roles and decided it’s for you; what do you need to learn and how do you sell yourself? If I was to go back in time and try again to move into data science again, I’d would still follow the same high-level path I took before, with some changes on what I spent the most time on:

  1. Learn the required skills
  2. Complete some personal/in work projects to exhibit those skills
  3. Advertise those projects to prospective employers

I’ll go through each of these in order in some more detail below.

Learning the skills you need

There are a wide range of skills that data scientists need to have, some of which you need to excel at, while knowing the basics of others can suffice. Knowing which of these is which can be tricky.

Before picking which resources you want to use to learn, consider what medium you best learn through. I find personally I learn best through textbooks and visual media, but collections of videos seem to be very effective for others. Additionally, be wary of falling into the trap of using bite-sized pieces of knowledge to give yourself the illusion of learning, when you’re really just skimming the surface of the required knowledge and missing deeper connections. Andrej Karpathy described this very well in this recent post on X/Twitter, from which I highlight a paragraph that I find quite impactful:

So for those who actually want to learn. Unless you are trying to learn something narrow and specific, close those tabs with quick blog posts. Close those tabs of “Learn XYZ in 10 minutes”. Consider the opportunity cost of snacking and seek the meal - the textbooks, docs, papers, manuals, longform. Allocate a 4 hour window. Don’t just read, take notes, re-read, re-phrase, process, manipulate, learn.

     Andrej Karpathy, Twitter/X

Below I describe roughly the order I would undertake learning, assuming a prior degree in some sort of science. Many of these resources are described in my page on data science resources.

  • Learn software development:
    • with the missing semester course from MIT.
      • this might feel like a dry one to start off with, but since this is probably the most fundamental skill here and all others rely on it, it is a good place to start. Being comfortable with you terminal and being able to collaborate using Git will make future learnings and your eventual job much easier.
  • Learn python:
    • I don’t have an exact python course I would recommend and given there are many different levels you may be starting at, I would steer you towards the choice of python learning paths on Real Python.
    • The main python website python.org has also indexed many useful resources
      • Out of all the skills here, python is the one you probably want to put the most time into being good at, because it will be the end product you will be producing a lot of the time.
  • Learn statistical modelling:
    • The book “An Introduction to Statistical Learning” is available with examples in R or Python here.
      • I’ve found this book to cover the level of detail needed to understand modelling as a data scientist in a very digestible way.
  • Learn how to do a machine learning project:
  • Learn (some) cloud computing:
    • How much you need to know will vary role by role, but understanding the fundamentals is good.
    • I don’t have a stand-out recommendation for learning cloud computing, but have had a good experience learning with Udemy’s courses on AWS in the past. This one may be a good introduction for those new to AWS or cloud computing.

After (or maybe while) you have learnt the above skills, you can begin to undertake some personal projects.

Doing some personal projects

You need to show that you can use your skills to solve problems and explain why you’ve made the choices that you have. Picking a good personal project can be tricky; you want something that is complex enough to challenge yourself, but not something that isn’t fundamentally unfixable. Remember that training a good machine learning model is only one of the components of providing a business-ready solution - if you are able to implement an end-to-end solution, possibly deploying in the cloud, that can be very impressive (but do remain vigilant of AWS costs piling up rapidly!).

Do something that is either relevant to an industry you’re interested in or that you’re personally interested in. This makes it more likely the problem and solution will be more relevant to roles you’re looking for and more interesting all round. Alternatively, if you can apply some data science to your current role, that is an excellent way to show you can take initiative and solve problems.

Finally, you want to make sure that potential recruiters will see these projects, and you can set yourself apart from other applicants. Put the code that you have developed in a public GitHub repository and make sure it has a story behind it - maybe you want to write a blog about it, or make a really interesting readme for the project, or find some way to host it online for others to use! Whatever way you decide to advertise it, make sure it clearly shows how you’ve solved your problem and links to your code. Additionally, do spend time tuning your CV to sell yourself and make sure it’s specific to whatever job you are applying to.

Conclusions

That wraps up some pointers for how I would go about trying to get into data science if I had to again. Bear in mind this guide is very much not end-to-end and many of the soft skills that are essential for a data science are not given the time they are due in this post. Some resources below include alternative opinions that might be useful.

Resources