data science, to begin with well done.
You’ve chosen probably the most lucrative and fast-growing careers in tech.
But here’s the reality: most students waste months (even years) spinning their wheels on the incorrect things. Avoid these mistakes to fast track your data science profession.
After 4+ years working in the sector, I’ve seen exactly what separates those that land their first data science job fast… from those that never make it past infinite tutorials.
In this text, I’ll break down the five biggest mistakes that hold beginner data scientists back so you’ll be able to actively avoid them.
Not Learning Fundamental Maths
Maths is by far an important… and yet also essentially the most neglected.
Many individuals, even practitioners, think that you simply don’t have to know the underlying maths behind data science and machine learning.
You might be indeed impossible to perform backpropagation by hand, construct a choice tree from scratch, or construct an A/B experiment from first principles.
So, it is straightforward to take this without any consideration and avoid learning any of the background theory.
Sure, you’ll be able to construct a neural network with a couple of lines of PyTorch, but what happens when it has weird behaviour and you must debug it?
Or what if someone asked you what the prediction interval is around your output from a linear regression model?
These scenarios come up more continuously than you think that, and the one way you’ll be able to answer them is by having a solid grasp of the underpinning maths.
Consider maths because the operating system of your brain for data science. Every model, every algorithm, every insight you produce runs on it.
In case your OS is buggy or outdated, nothing else runs easily, irrespective of how fancy your tools are.
Lay the foundations now when you are in the training phase, as this may help you move much faster later in your profession.
Trying To Find The “Best” Course
I often get asked:
I actually do love you all, but this query must go away.
As a whole beginner, one of the best course is the one you select and complete.
Many introductory courses in data science, machine learning, and Python will teach you an identical things.
Chances are you’ll discover a teacher or a teaching style higher than one other, but fundamentally, you’ll acquire very similar knowledge to a different person doing a little other course.
Bias towards motion and getting entering into the start, you’ll be able to later adjust your direction for those who feel you’re misaligned. Stop overthinking.
Because the famous saying goes:
Everyone’s journey and background are different, and there isn’t a “a technique” to interrupt into data science.
So, take everyone’s advice (even mine) all the time with a pinch of salt and tailor it to yourself. Do what feels right and best for you.
Not Doing Project-Based Learning
Along that theme, one other common pitfall is tutorial hell.
When you are unaware of what tutorial hell is, this blog post explains it thoroughly:
You might be principally following tutorial after tutorial and never attempting to construct anything on your personal.
To learn the concepts, you must practice and apply them independently in your work. That is the way you solidify your understanding, and the training is completed.
Imagine that you’ve got only ever built an XGBoost model following online tutorials.
When you are then given a takeaway case study as a part of an interview, you’ll really struggle as you’ve got had no experience constructing models with out a step-by-step walkthrough.
What I advocate for is “project-based learning.”
Trust me, this approach is exponentially higher than doing quite a few tutorials (speaking from painful experience here!).
Quantity Over Quality Projects
Whilst doing projects is one of the best technique to learn, don’t oversaturate your GitHub with a great deal of “easy” projects.
If all of your projects revolve around an already pre-made dataset from Kaggle and using sci-kit learn’s .fit() and .predict() methods, it’s probably time to try something a bit harder.
Now, I’m not slating these entry-level projects, as they’re a terrific technique to get your hands dirty.
Nonetheless, sooner or later, the standard of your projects will matter greater than the amount.
Larger, in-depth projects will likely be those that really get you hired. Recruiters don’t need to see one other titanic dataset problem; if anything, it might be a red flag nowadays.
Some ideas to try:
That is in no way an exhaustive list, and one of the best project is the one which is personal to you, as I all the time say.
Jumping Straight To AI
I’m going to be honest with you.
I’m an AI hater.
The explanations I’m not apprehensive could fill an entire video, so I’ll leave that for later. Nevertheless it’s actually funny, almost how little I’m concerned by it.
Anyway, the rationale I say that is that it baffles me once I see beginners jump straight into learning AI and LLMs.
That is a primary example of shiny object syndrome.
As a beginner, deal with the fundamentals of maths and statistics, and on old-school algorithms resembling decision trees, regression models, and support vector machines.
These are evergreen and can remain around for a very long time, so it’s clever to speculate in them early on.
AI remains to be an unknown entity, and whether it would be as popular and helpful in a couple of years is tough to inform.
If the subject is popular now and indeed helpful, it would be popular 1 yr, 3 years, and even a decade from now. So, don’t worry, you’ve got loads of time to check cutting-edge topics.
Remember what I said earlier about not all projects getting you hired?
That longer, more in-depth ones make all of the difference?
Well, see my previous article, which walks through specific projects that make it easier to stand out (and which of them are a complete waste of time).
See you there!
One other Thing!
Join my free newsletter where I share weekly suggestions, insights, and advice on landing your first data science or machine learning job. Plus, as a subscriber, you’ll get my FREE Resume Template!
https://newsletter.egorhowell.com
