Hiring is one of the hardest parts of building a startup, and it gets harder the better the people you want are.
At a startup, you are almost always competing against big tech for the same talent. You can’t match their compensation. You sometimes lose people to them anyway. And you have to keep a team motivated while you fight through the trenches, where the battles and the longer wars can feel relentless. None of that goes away. But over the years I’ve learned that the constraint forces you to get better at the parts of hiring that actually matter, and to look for talent in places most companies never bother to check.
Here’s what has worked for me.
Look for a Track Record of Beating the Odds
The single most useful framework I’ve adopted comes from Who: The A Method for Hiring by Geoff Smart and Randy Street. The idea I keep coming back to is to look for adversaries a person has overcome in their life, without ever crossing a legal or ethical line to do it.
A story makes this concrete. I was once interviewing a candidate and asked how many years of coaching he took before sitting for the IIT entrance exam. He said two. I knew that most people in India spend four years on this, essentially all of high school from 9th to 12th grade. So I gently pushed on the gap. It turned out he had only two years of coaching for financial reasons, and despite that he secured a rank around 200 out of roughly 1.2 million people who took the exam.
That answer told me more than any credential could. We kept going and talked through other adversities he had overcome later in life, including in graduate school. People who have repeatedly found a way through hard circumstances tend to keep doing it, and that is exactly the kind of person you want in the foxhole with you at a startup.
Watch How People Work When No One Is Grading Them
Some of the best signal shows up after the interview, not during it.
I once hired a candidate who did not do well in about half of his interviews. What changed my mind happened on its own. Over the Christmas break, without anyone asking him to, he went back and completed all of the assignments based on the interviewers’ feedback, and then did a few more on top of that.
I hired him, and he turned out to be one of the best data engineers I have ever worked with. The interviews measured where he was that day. His response to the feedback measured something far more important, which is how he works when no one is watching.
Finding Talent in Unusual Places
This part applies whether you are at a startup or a large company, but it matters most when you cannot win on compensation alone. When you can’t pay what big tech pays, your network is the only reliable way to reach A players.
Referrals. This has been my single best source. I have consistently found that people will take a risk, work for less money, and take a leap of faith for someone they trust. It is worth investing in. Incentivize referrals with a larger than average bonus, and protect your team’s experience so your internal NPS stays high. A great net promoter score among your own people pays for itself many times over in who they bring with them.
Tech talks and podcasts. When you come across a great podcast guest, invite them to give a tech talk. People almost always enjoy sharing what they know, and it gives you a natural way to share what you are building. It is a quiet, low-pressure form of hiring. For me, the most relevant show over the last seven years has been Flirting with Models. Even when a guest isn’t looking to make a move themselves, they usually know many other people working in the same domain.
Big tech layoffs. When large adtech organizations went through layoffs, places like Amazon’s Twitch and Meta’s adtech teams, they became some of the best places for us to find genuinely strong ML researchers and data engineers. Talented people suddenly become available through no fault of their own, and a startup that moves quickly can do very well here.
Open source and competitions. Contributors to the repositories you depend on are a great pool, for example foundational time series models, as are people who do well in Kaggle competitions. Their work is public, which means you can evaluate real output before you ever talk to them.
Professors. Ask professors in your domain for a list of their strong students. Faculty often have a clear read on who is exceptional, and they are usually happy to point you toward people who are about to enter the market.
The Constraint Is a Gift
You will not win the compensation war against big tech, and you don’t have to. What you can do is hire for grit over polish, pay attention to how people behave when no one is grading them, and build a network so strong that the right people come to you.
I have found that the people who join a startup for reasons other than the biggest paycheck tend to be exactly the ones you want when the work gets hard. The constraint forces you to look for them, and looking for them is what makes the team great.