The person who became my best hire barely spoke for the first hour.
TL;DR
- Hackathons reveal how people think under pressure — interviews reveal how they present under preparation
- Twenty minutes isn't enough time to fake depth of thinking
- Screen for connective thinking and ego-resilience, not credentials or domain expertise
- The best hires come from processes that strip away the performance layer
- Rejected candidates can become a powerful freelance network if you treat them well
We were at IIT Bombay. Not for a placement drive — I'd fought the placement coordinator for weeks to do something different. He told me nobody had ever run a hackathon as a hiring exercise. That they didn't have the logistics. That it wasn't how things were done.
I didn't fully know how I wanted to do it either. I just knew that a group discussion — twenty minutes of polished talking — would tell me nothing about how someone actually thinks under pressure. And thinking under pressure was the only thing I cared about.
Round one: Controlled chaos
I gave them a real business problem. Not a case study — something I was genuinely dealing with at Welspun One. Then I said: you have twenty minutes to think, then you pitch a solution using whatever technology you know.
The room was a mix. Geospatial ML researchers. Industrial AI specialists. Computer vision people. I didn't want to filter by domain. I wanted to see how well they understood their own subject matter and — more importantly — whether they could connect it to an actual business outcome.
Most candidates in a traditional interview will tell you they can do this. A hackathon shows you whether it's true.
Twenty minutes isn't enough time to fake it. Some people froze. Some defaulted to textbook answers. And a few — the ones I was watching for — started pulling their own knowledge apart and rebuilding it around the problem in front of them. You can't teach that. You can only create the conditions for it to reveal itself.
Round two: Where the real signals emerge
For the second round, I put them into groups of three or four and gave them a new problem. This time I explained the full lifecycle of our business: we acquire land, we construct, we lease, and then a fund exits the asset. Pick any part of that lifecycle. Build a solution. You have thirty minutes.
I did this before lunch on purpose. They were hungry, tired, and couldn't hide behind preparation.
What happened next was the whole point. I wasn't evaluating solutions. I was watching structures form in real time.
Some people immediately took charge — not because they had the best idea, but because they couldn't tolerate ambiguity. Others defaulted to execution, heads down, building whatever they were told. A few tried to mediate between competing ideas. And one person — quiet through most of round one — started doing something I didn't expect.
She wasn't the loudest. She wasn't pitching a vision. She was listening to the three competing ideas in her group, finding the connective tissue between them, and pulling the team toward a coherent output before time ran out. She was the one who made sure they actually delivered something.
That's the person I hired.
What I was actually screening for
I went to IIT Bombay for a reason. I knew these students had already been through one of the most intense academic filters in the country. IQ wasn't the variable. Discipline, grit, the ability to push through hard things — that was already proven.
What I needed to see was something LinkedIn profiles and resumes can never show you: how someone behaves when the structure disappears. When there's no rubric, no right answer, no time to prepare. Do they organize or do they wait? Do they connect ideas or do they protect their own? Do they care more about being right or about the team delivering?
I wasn't looking for the smartest person in the room. I was looking for an empty vessel with the right instincts — someone I could shape, who was ready to learn, and who wouldn't break when things got ambiguous. Because building AI inside a traditional company is ambiguous every single day.
The solution they proposed didn't matter much. What mattered was what I saw them do.
Version two: The design hackathon
A few months later, I needed to hire for a completely different capability — visual design for AI-generated outputs. This had nothing to do with technical skill. It was about aesthetic instinct, about understanding how AI tools could be directed toward a specific visual standard.
So I built a second hackathon, this time as a marketing funnel.
First, we cast a wide net on LinkedIn (yes, the irony — LinkedIn is great for sourcing, terrible for evaluating). About fifty people applied.
Then the filtering began. The first step was an NDA. This might sound trivial, but it was the most effective filter in the entire process. The candidates who balked — who felt that signing an NDA for a "tryout" was beneath them — self-selected out. What looked like an administrative step was actually a test of ego. The people who signed without drama were telling me something important: they cared more about the opportunity than about protecting their status.
From fifty, we got to about twenty who signed. Then we asked them to submit a short video using a specific AI toolkit we recommended. This filtered for follow-through and the ability to learn a new tool quickly. About ten to twelve made it through.
The final round was supposed to be in person, in Mumbai. We had the logistics planned for a physical session in Lower Parel. But then something happened that changed the whole process — candidates from Brazil and from tier-2 and tier-3 cities in India asked if they could join remotely. We didn't have the infrastructure for it. We hadn't planned for virtual participation at all.
But they were so excited, so genuinely invested in the process, that we couldn't bring ourselves to say no. We figured it out on the fly.
The person I hired from that cohort was one of the remote participants.
Friends of AI Labs
Here's the part nobody plans for.
Out of the twelve people who made it to the final round of the design hackathon, I could only hire one. In a normal process, that's where the relationship ends. You send a polite rejection email and never think about those people again.
But something different happened. These candidates had been through an experience that was genuinely meaningful to them. Multiple people told me — during the hackathon itself — that regardless of whether they got the job, this was the best hiring experience they'd had. They felt seen, challenged, and respected.
So when it was over, they didn't disappear. They stuck around.
We started calling them Friends of AI Labs. It wasn't a formal program. It wasn't planned. It was just a name for what had already happened organically — a group of talented people who wanted to stay connected to what we were building.
Three months later, when we had a design requirement, it took one phone call. They remembered us as the design hackathon team. They were ready to work immediately.
Today, I have four to six freelancers from that original cohort who work with us part-time, after their regular 9-to-5 jobs. This is their passion project. They give us their best work not because we pay the most — we pay fairly, but we're not outbidding agencies — but because they genuinely want to be part of what we're building.
I feel a social responsibility not to take advantage of that. These aren't people to exploit with below-market rates because they're enthusiastic. They get fair pay for the four to five hours a week they contribute. But the relationship is built on something that money didn't create and money can't replicate.
The system underneath
When I zoom out, everything I've described follows the same pattern.
The IIT Bombay hackathon filtered for depth of thinking over surface presentation. The design hackathon funnel filtered for commitment and ego-resilience over credentials. The Friends of AI Labs network formed because we treated candidates like humans, not like pipeline.
None of this happened on LinkedIn. LinkedIn sourced the initial candidates — it's excellent at that. But every meaningful signal, every insight into who someone actually is and how they actually work, came from a process designed to strip away the performance layer and see what's underneath.
I apply this same philosophy to everything. My website exists instead of a LinkedIn presence because I'd rather be found by people who go deeper than be visible to people who stay on the surface. My AI department was built by people who proved themselves under pressure, not people who interviewed well.
Depth is a strategy. Most people treat it as an inconvenience.