A year ago, I dropped out of grad school to co-found a startup with my brother. Our goal was simple enough: fix the data science talent shortage.
Since launching, we’ve placed over 60 brilliant new grads and developers with great companies. And we’ve learned a lot along the way.
One thing we noticed early on was that the talent shortage disproportionately affects new startups. It also disproportionately affects enterprise companies that don’t have much AI expertise to start with. Why is that?
Another way of asking that question is: what do AI-savvy companies do differently?
When we dug deeper, we found that the world’s top tech companies tended to have one thing in common in their approach to sourcing data science and deep learning talent: a massive focus on campus recruitment.
A huge fraction of the machine learning engineers that Facebook, Google and Airbnb hire are junior developers and new grads. The tech giants catch them before everyone else can, scooping them up early with internships and work terms.
There are mathematical reasons why this approach outcompetes the others. But for now, let’s think about 1) why other companies aren’t already doing this; and 2) why it’s causing them to miss out on great prospects.
Why some companies insist on experience
If you look up data science or machine learning jobs on Indeed, you’ll find an ocean of postings that ask for “a PhD in machine learning or related field with 3+ years’ experience.”
Let’s pretend that PhDs in machine learning with 3+ years’ experience spend any time on Indeed (they don’t). Let’s also pretend that the average company has a chance of standing out to them in Indeed’s sea of job postings (it doesn’t).
That still leaves the question: Why a PhD? Why not a Masters’ degree? And why not 2 years’ experience? Or 18 months?
Here’s an interesting fact: in reality, there aren’t a lot of cases where the work of an ML engineer actually requires them to have a PhD. So what’s the real point of asking for one?
The real point is that a PhD is a decent proxy for skill, if you’re not sure what actual skill looks like.
AI-savvy companies like Facebook, Samsung, and Apple don’t require PhDs. They hire inexperienced talent all the time because they’re confident in their ability to find, identify and screen that talent. They know their tech well, and they’ve seen enough interns that they can correlate their performance on coding tests and technical interviews with their raw ability and potential.
In other words, AI-savvy companies have enough experience with talent discovery that they don’t need to use formal education, or years of experience, as a proxy for skill.
For almost everybody else, the “PhD with 3+ years’ experience” is just a proxy for technical ability and learning potential. It feels like a safe bet because someone else — a university or a previous employer — has vetted the candidate for you.
But shyness about hiring new developers and graduates has a painful side effect.
What those companies are missing
On the face of it, it seems clear what companies that aren’t Facebook or Google need to do to catch Facebook- or Google-level talent. They need to find that talent the way Facebook and Google do.
For most companies, that means pushing campus recruitment. If the top startups in Silicon Valley are picking off fresh new graduates (and even undergrads), and you’re holding out for PhDs with 3 years’ experience, then you won’t even get a shot at the unicorn hire that you really want.
Early talent discovery is key. But to discover talent early, companies need two things. They need a way to reach out to new grads who have the right technical skills; and they need a way to screen them accurately.
Most companies that aren’t already tech giants lack one or both of these. They also don’t have the time to build and tweak all the infrastructure they’d need to attract and assess new grads with confidence. Even a Silicon Valley giant sometimes misses promising people, or hires a dud.
But that means companies are missing out on the world’s best talent. What’s more, they’re filtering for candidates who have been assessed by the market (through prior experience), but who are also looking for work.
That’s a dangerous combination, because most companies try hard to hold on to their best employees.¹ That means the very best hires are rarely looking for work, unless they’re new grads.
What we’ve learned
We started SharpestMinds to level the playing field between the Silicon Valley giants, and the companies that can’t afford full scale campus recruitment and screening operations. We wanted to find a way to reach Google-level talent before Google did.
Just as importantly, we wanted to give great new grads and experienced developers a shot at the big leagues in data science and machine learning.
To do that, we’ve always been background-agnostic. We don’t care what degree you have, as long as you can pass our coding tests and technical interviews.
An interesting consequence of this is that we’ve ended up with an eclectic collection of members. Some of them are PhDs in machine learning, or former interns at Google, Apple, and Tesla. But we’re regularly surprised by physicists, mathematicians, and even economists, who are fully self-taught, but who are in the same league as the Googlers.
After parsing thousands of applications and interviewing hundreds of candidates, you get good at spotting the ones who are massively undervalued by the job market. In fact, the most important thing we’ve learned in the past year is just how many of them there are.
¹People become unemployed for all sorts of reasons, of course, so I’m not saying that someone who’s unemployed must be unemployable. The point here is rather that the pool of talent that has been assessed by the market and also unemployed will be enriched for low-quality talent (statistically).