You Don’t Need to Be a Math Genius for Data Science

Learn the truth about data science interviews and how deep you actually need to go into maths.

Memorizing every math theorem won't get you through the data science interview.

You think it will? Well, you're in for a big surprise and wasting months of your life.

Here's a dirty little secret. Most interviewers don't give a flying f*** about you being able to derive eigenvalues. 

So, if not math, what do you need to study to crush the interviews?

The Truth About Data Science Interviews

Here's a reality check. Most data science and machine learning engineer interviews are not about proving you're a math god. Fact. 

The interviewers care about three things. 

1. Do you understand the concepts? 
2. Can you apply them? 
3. Can you talk about them without sounding like you swallowed a math textbook?

4 Core Topics You Need to Master

The topics you need to know inside out are these. 

1. Probability
2. Statistics
3. Linear algebra
4. Optimization (bonus)

What’s Overkill (Stop Wasting Time!)

Unless you specifically apply to roles that scream “PhD math nerds only”, you're unlikely to be asked “Find the posterior predictive distribution for a Beta-binomial model” or "Explain the Shannon-Nyquist theorem... using interpretive dance." 

Instead, you'll probably hear “When would you choose mean absolute error over mean squared error?”, or “How does principal component analysis work in human words, please?” 

Pro tip: Interviewers love candidates who can explain complicated stuff simply. 

Like explaining eigenvectors to a 5-year-old. 

(Hint: It’s just directions that don’t change when you squish/stretch things!)

Unpopular Opinions on Data Science Prep

One, 90% of ML math difficulty is wildly exaggerated. (Sorry if you thought this was a quant role. It's not.)

Two, some interviewers don't even ask math questions at all – they just shove SQL and LeetCode at you. 

Three, if you name-drop a model in the interview, you'd better know it. (Mention Expectation-Maximization and fumble? You're toast.)

Translation: Don't flex what you can't defend. 

The Ultimate Preparation Plan

If you're balancing work and prepping, follow these steps.

1. Master the top five concepts from each core subject.
2. Practice explaining algorithms like KNN, random forests, logistic regression.
3. Practice implementing the algorithms, e.g., LeetCode problems, StrataScratch algorithm questions and projects, Kaggle datasets. 
4. Know your resume. If you list PCA, you'd better know why it exists. 
5. Understand model assumptions. Bonus points: Explaining model assumptions from a project you've done in the past (because real data ≠ textbook examples). 
6. Optional: System design for machine learning engineer interviews, i.e., how do you scale or deploy a model without causing the servers to melt?

Conclusion

So, do you need some math for DS/MLE interviews? Absolutely.

Should you spend 400 hours proving the Lebesgue integral from scratch? Only if you enjoy existential crisis. 

You should prepare smart, not masochistic.