What kinds of technical questions do they ask in a data science interview?

On the stats side, often higher-level theory questions, such as "How does the k-means algorithm work?", "How do you select the best k for k-means?", "What is the curse of dimensionality?" which again would not be things covered at a data boot camp or data science thought pieces on Medium.

On the technical side, there is often more-advanced SQL (nested JOINs + PostgreSQL window functions). On the big data side, there is often discussion of distributed systems (e.g. Spark clusters) and practical algorithmic complexity at scale (i.e. instant fail if you suggest anything loglinear or slower).

I overheard some colleagues talking about a recent interview where a candidate with "stellar industry experience", i.e. Kaggle wins and previous ML experience at a valley company, who couldn't explain Bayes rule to them, let alone rederive Naive Bayes. While books like the one below are extremely theoretical, anyone interviewing for these kind of roles should spend at least a week or two just looking through this to see what kind of algorithms and properties are studied in theory.

Foundations of Data Science (Blum, Hopcroft, Kannan) https://www.cs.cornell.edu/jeh/book2016June9.pdf

This is like not hiring a [big name coding competition] winner because he didn't know radix sort.

I think it's more like not hiring a big name coding competition winner because they never bothered to learn how to use version control, or any coding best practice, or any language other than C.

Trying to do data science with zero knowledge of the fundamentals of probability is dangerous. Bayes rule isn't some kind of deep magic, it's covered within the first few lectures of an undergraduate probability course and it's absolutely necessary to understand the output of any machine learning model.

>I think it's more like not hiring a big name coding competition winner because they never bothered to learn how to use version control, or any coding best practice, or any language other than C.

Depends on what you're hiring for, but I'll take "competition winner with no version control" over "average programmer with expert VC capabilities".

>Bayes rule isn't some kind of deep magic

Yes, it's largely conceptually obsolete.

The people jamming out weekly SOTA machine learning models on arxiv aren't sitting around meditating on conditional probabilities. They're making little tweaks to giant models that are basically impossible for a human to comprehend.

> Yes, it's largely conceptually obsolete.

I'm sorry, what? How did you arrive at a point where you believe this is true? This is like calling compilers "obsolete."

Is it because you believe deep learning has "taken over" or something?

Try to derive e.g. a face detector from bayes theorem. You immediately arrive at computationally intractable sums/integrals. Yet, we have super-human image classifiers. Therefore, bayes theorem is obsolete. Sure, you can try to retrofit bayes theorem on top of a neural net, but who cares?

> You immediately arrive at computationally intractable sums/integrals.

So we instead sample from that posterior.

Unless you think MCMC is also obsolete, in which case I’ll see myself out.

You're right, but a) you have comp efficiency issues with MCMC, and b) just empirically MCMC models don't work as well as gradient descent + NN for many tasks.

And you don't have computational efficiency issues with NNs?

We're also ignoring the benefits of a posterior distribution, which is useful for understanding the data-generating process.

Yeah of course. I can't explain to you why NNs outperform bayesian approaches, probably just NNs are capturing the correct type of prior for vision/language tasks. And yeah bayesian models are more interpretable but when you have millions of latent variables I'm not sure interpretability is a thing.

Yep, we arrived at my larger point: if you care about interpretability, NNs are horrible and Bayesian techniques are pretty damn great.

Well certainly, but interpretability is obsolete.

Now you're just trolling. :)

I'm not sure I agree with that. I don't know any ML researchers that don't know about probability, but maybe they exist somewhere. Machine learning research isn't a good model for "data science" writ large.

Maybe there are some jobs and some problem spaces where you can just tweak big black box models and you don't ever need to think about what their output means. But if you're the kind of data scientist who helps make decisions with data -- you better believe statistics and probability is conceptually relevant. As soon as models meet the real world, you've got to understand probability in order to know what to expect.

> Yes, it's largely conceptually obsolete.

Wow

All the stats stuff you mentioned (and more) is laid out very nicely in the second chapter of "Elements of Statistical Learning" https://web.stanford.edu/~hastie/Papers/ESLII.pdf

Having never been to a data boot-camp -- what _do_ they teach there, if those wouldn't be covered?

Galvanize, the one mentioned in the article and markets the most, is vague on their curriculum but is more scikit-learn oriented. https://www.galvanize.com/san-francisco/data-science#curricu...

Metis, another known bootcamp, does explicitly mention things like k-means, which is something I didn't know: https://www.thisismetis.com/data-science-bootcamps

Having spoken to several students as well as looking at their GitHub projects from the galvanize data science program, I am fairly confident they cover those stats subjects in question.

Not to say that a data science boot camp is all you would ever need to know, but I would give them a little more credit in what they teach. That galvanize link is not really their syllabus, they probably keep it vague to force you to sign up to their email list to get the real one.

I'm not sure which bootcamps you're referring to, but I went to one that went into the topics from your first sentence in depth. The topics from your second sentence were not as thoroughly explored, though.

The stat questions you mentioned are all answered from a introductory modeling book or class. I am surprised those subjects are not even talked about in a data boot camp.

Hi, minimaxir. If confidentiality doesn't prohibit it, I'd be curious to know what the most interesting roles you interviewed for were.

If you had to summarize into three points?

I have one notable observation:

On Hacker News, every time an interview thread pops up, there is a discussion decrying the use of technical screenings before an onsite, and often suggest practical work experience instead using a homework assignment (which this article does not discuss).

Most of the companies I've talked with for data analyst/science roles have given me both a homework assignment and a technical screen before the onsite. And often a prescreen test before both of them.

There have been a number of occasions where I easily passed the homework screen but failed the technical screen (without any feedback as to why). And it's beginning to get annoying.

Did you find another job yet? If not, what type of opportunity are you looking for? Data science is also a wide, somewhat poorly defined domain.

Still looking. Mostly for any role with a data analyst/data scientist title (i.e. I am not applying for the machine learning/NLP roles which require a PhD and the authorship of several papers since there is no point).