In 2025, 1,80,422 candidates sat JEE Advanced, 54,378 qualified, and the 23 IITs carried a sanctioned intake of 18,160 seats — so only about one qualifier in three actually got in. That exam is itself gated: sitting it at all requires clearing JEE Main, which 14,75,103 candidates attempted across the 2025 sessions. If you are a seed-stage founder deciding where to look for engineering interns, that funnel is the single most useful thing to understand — not because it tells you who can write software, but because it tells you the expensive part of narrowing has already been done, and somebody else paid for it.
Key takeaways
- The admission funnel is a free narrowing step — but it measures an exam sat at seventeen, not shipping software. It buys you a smaller pool, not a verdict.
- Your real cost is evaluation. Startups under 25 people read 298 applications per hire — roughly ten hours of founder reading before anyone talks.
- Artefacts have stopped settling the question: 35% of hiring managers have had someone other than the applicant show up to a virtual interview.
- A six-minute code walkthrough replaces most of the first two rounds. It is the one artefact that is hard to fake and fast to judge.
Your bottleneck is evaluation, not sourcing
Founders describe intern hiring as a sourcing problem — where do I find good people — and then spend their time on something else entirely. Ashby’s 2026 State of Startup Hiring, built on 11 million applications and 32,000 hires across more than 1,200 venture-backed companies, found that companies under 25 employees read 298 applications per hire. Sourcing was never the constraint. Reading was.
At two minutes per application — generous for a CV, stingy for a portfolio — that is close to ten hours before a single conversation, and at seed stage those hours come out of the founder’s week. This is the cost people mean when they say hiring distracts from customers. It is not the interviews. It is the reading that precedes them, almost all of it spent on candidates who will not be hired.
The artefacts you were reading stopped working
The standard fix is to stop reading CVs and start reading work: portfolios, repositories, deployed projects. That fix is now failing on its own terms. Checkr’s 2025 survey of 3,000 US hiring managers found that 31 percent had personally interviewed someone using a fake identity and 35 percent confirmed that a person other than the listed applicant took part in a virtual interview, with AI-generated fake work samples and portfolios the single most-cited form of deception. Nearly six in ten suspected candidates of misrepresenting themselves with AI at all.
So the repository is no longer evidence. It is a claim, and verifying it costs you the same hour a CV always did. Every cheap filter in the funnel has been commoditised at once, which is why the founder’s screening load keeps rising even as the tools promise the opposite.
What the admission filter does and does not buy you
Be precise here, because the sloppy version of this argument is easy to sell and wrong. The IIT, NIT and IIIT funnel is a real filter, applied at scale, by institutions with no incentive to flatter anyone. It selects for sustained preparation, problem-solving under time pressure and a willingness to grind at something difficult for two years. Those traits are not nothing.
What it does not do is tell you whether a particular nineteen-year-old writes maintainable code, reads someone else’s system without panicking, or says “I don’t know” when they don’t. The exam was sat at seventeen and measured none of that. Anyone selling you the pool as though admission were a proxy for engineering ability is selling you a shortcut that does not exist. The correct use of the funnel is narrower and more useful: it collapses the size of the pool you have to evaluate, so that the evaluation you still have to do is affordable.
A six-minute evaluation loop
- Ask for a code walkthrough, not a portfolio. Five to eight minutes of screen recording: their editor, their code, their voice. Make it the first thing you request, before any CV.
- Specify the hard part. “Show me the piece of this that gave you trouble and tell me what you tried first.” That single instruction does most of the filtering for you.
- Watch ninety seconds before deciding to continue. Someone who wrote the code navigates it from memory. Someone who did not describes features instead of files. The difference is obvious and it is fast.
- Listen for the discarded approach. The decision they rejected tells you more about their judgement than the one they shipped.
- Reward the admitted weakness. A candidate who names what is still broken in their own project is showing you the self-assessment you will depend on when they are working unsupervised.
- Interview only the ones you would work with. The walkthrough has already covered authorship, reasoning and communication, so your live time goes to fit and scope rather than verification.
What you can delegate, and what you cannot
Founders reach for interns hoping to buy back their week, which is a reasonable goal pursued in the wrong place. The hours you can genuinely reclaim are the screening hours above, plus well-scoped implementation work: an integration nobody has got to, the test coverage on a module you already understand, the internal tool three people keep asking for. That work is real, it ships, and it does not require you to be in the loop hourly.
What you cannot delegate is the review. A study of 45,865 first contributions across 777 GitHub repositories found that how maintainers interact with a newcomer predicts long-term retention better than the properties of the task itself — we went through that research in detail here. An intern with a short review loop compounds. An intern left alone with the product because the founder has moved to full-time selling produces work nobody trusts and nobody merges, and the six weeks are gone. Keep the review; give away the screening.
That is the honest version of “focus on customers”: not stepping away from engineering, but stopping the ten hours of reading that never told you anything a six-minute recording would have told you in the first ninety seconds. We ask every student on Tierones for that recording, and it is the first thing we show you — the reasoning behind it is here.
FAQ
Why hire interns from IITs, NITs and IIITs specifically?
Because a large selection step has already happened at no cost to you: 1,80,422 sat JEE Advanced in 2025 against a sanctioned IIT intake of 18,160. What it does not tell you is whether any individual writes good software — it measures an exam sat at seventeen. Use it to shrink the pool, then evaluate properly.
How much of my time does intern hiring actually take?
Startups under 25 employees read 298 applications per hire. At two minutes each that is about ten hours of reading before a single conversation — and at seed stage that reading is nearly always done by a founder.
Can I hand product work to interns and step back from engineering?
Not safely. Research across 45,865 first contributions found that maintainer interaction predicts whether a newcomer sticks better than the task itself does. You can reclaim the screening time and the well-scoped implementation work. You cannot reclaim the review.
Tierones works with students from IITs, NITs and IIITs who present their work as a code walkthrough rather than a CV. Tell us what you need — we’ll show you proof, not CVs.