Careers · 5 July 2026 · 6 min read

The hardest part of building engineers, from the hiring side

Every programme claims to build engineers. From the hiring side, most don't. Here's what the gap actually is, and why it isn't the curriculum.

In this article

I’ve been on the hiring side of engineering for a while. At Google, at HackerRank, and now at Kalvium. Across all of it, one pattern stays consistent: most programmes produce people who know engineering, not people who can do it. The gap between those two things is the hardest part of building engineers, and it’s the part most programmes never actually address.

The assumption everyone makes

The assumption is about curriculum.

Talk to anyone who runs an engineering programme and they’ll say the hard part is keeping the subjects current. The tools need to match what companies use, the faculty need to understand the industry, the electives need to track the market. It’s a reasonable thing to worry about.

It’s also not where the difficulty is.

The candidates who come through my hiring loops almost never fail because their curriculum was wrong. They fail because they’ve never actually worked. They’ve studied. They’ve practised. They’ve passed exams on the subject matter. But they haven’t done engineering in a real environment with real constraints, under conditions where something would actually break if they got it wrong.

What “working” means from the other side of the table

Here’s what the gap looks like from where I sit.

Can you navigate a codebase you didn’t write, in under an hour, and tell me what’s wrong with the function on line 200? Not write a new function from scratch. Navigate an existing one. Most candidates who’ve spent four years building solo assignments from scratch have never had to do this in a setting where the answer matters. It shows immediately.

Can you own something that could break on real users? Not a homework assignment where the only consequence of failure is a reduced grade. Something where if it goes down, people notice and someone calls you. The relationship to that kind of ownership is fundamentally different. It changes how carefully you check things before shipping. Most candidates haven’t had it.

Can you decide what to do next when nobody’s telling you? The first week at a real job is mostly figuring out what you don’t know, then making a decision about where to start anyway. Candidates who’ve spent four years following prescribed syllabuses hit a wall here. The muscle of working without someone pointing the direction doesn’t exist until you’ve actually built it.

Three things. None of them is on a marks card anywhere.

Why most curricula miss all three

The structural reason is simple.

Every assignment in a typical engineering programme is isolated. You write it, you submit it, you get a grade. The code stops existing the moment the marking’s done. No one else reads it. No one else depends on it. Nothing breaks if it’s wrong after submission.

That feedback loop teaches you to produce code that passes the test. It doesn’t teach you to produce code that holds up, that other people can read, that handles what happens when something goes sideways after you’ve moved on to the next assignment.

Four years of that and the missing muscle is accountability to a real outcome. It isn’t missing because the students are weak. It’s missing because the programme never created the conditions where that muscle was needed. The students responded rationally to the incentives the system gave them: optimise for the exam, not for the outcome.

What production-from-Year-1 actually builds

Here’s the thing that changes the trajectory.

If you’re building something real in Year 1, where the code has to actually run, where someone reviews it as a colleague rather than an examiner, where the feedback is “this doesn’t handle the edge case, go fix it” rather than a grade out of ten, you build a different relationship to your work. You learn, early and repeatedly, that code is something other people will depend on. That changes the discipline you bring to it.

That relationship is what the hiring loop tests. Not whether you know what a binary search tree is. Whether you treat code like something that’ll outlive the moment you write it.

The specific stack doesn’t matter for this. A Year 1 stack will be outdated by the time someone graduates. What transfers is the habit: ship something real, have someone check it like it matters to them, fix what breaks, repeat. Do that for four years and you’ll arrive at a first job already knowing what the job asks of you, because you’ve been doing a version of it since Semester 1.

That’s what production-from-Year-1 builds. And it’s the thing most programmes skip because it’s harder to set up than lectures. You need real constraints, real code review, and feedback from someone who isn’t just awarding marks. Most four-year programmes aren’t structured for it.

What building for this looks like at Kalvium

I helped build Kalvium partly because I’d seen this gap from the hiring side enough times to know what closing it actually requires.

Kalvium students ship working software from Semester 1. Not a tutorial project. Software with real requirements and real code review. The DOJO runs daily coding practice six days a week. That isn’t just about grinding algorithms. It’s about building the habit of being uncomfortable, sitting with a hard problem, and working it out anyway, six days a week for four years.

From Semester 3, students build alongside industry partners. The feedback they get isn’t “well done, here’s your mark.” It’s “this doesn’t scale, rethink the data model” or “the error handling’s incomplete, it’ll break in production.” That’s the feedback that builds the right relationship to code.

In later years, students choose a production track: working in company-like sprint environments, interning with tech-first companies, contributing to open-source codebases, or building AI-native products. All of these create the accountability to real outcomes that most programmes never do.

The programme works across nine partner universities for Admission Year 2026-27, spanning Karnataka, Tamil Nadu, Andhra Pradesh, the NCR, Punjab, and Rajasthan. Students apply through the KNET, and admission includes a Psychometric Assessment, the KNET aptitude test, and an In-Person Interview. Selection isn’t just on technical knowledge. It’s on learnability and the capacity to be accountable to real work.

The 2026 batch was 82.40% placed as of March 2026, with a median of 16.5 LPA. Companies like Morgan Stanley, PhonePe, Thoughtworks, Lowe’s, and Yellow.ai are on the recruiter list. That outcome isn’t coincidence. It’s what happens when you close the gap between knowing and doing early enough to actually matter.

The one thing

The hardest part of building engineers isn’t the curriculum. It isn’t keeping the subjects current or finding the right faculty or tracking the latest frameworks. Those are real problems, but they’re solvable problems.

The hard part is creating the conditions where students have to actually work, under real constraints, accountable to outcomes that extend beyond the examination. That’s the discipline every hiring loop tests for. And it’s the one thing most four-year programmes still don’t build.


For a breakdown of what that hiring loop specifically tests, the companion piece on the six skills the branch name doesn’t mention covers each one with the interview patterns I see.

For the learning science behind why production-from-Year-1 works, the research on what engineering programmes that actually produce engineers do differently names the researchers and the specific mechanisms.

For families trying to apply this to an actual college decision, the framework for choosing a B.Tech CSE programme translates it into the five questions to ask before committing.


Anil is a co-founder of Kalvium and previously led engineering teams at Google and HackerRank. He runs hiring loops regularly and writes about what the Indian tech market actually rewards, from the interview side of the table. Read more from Anil or browse the Careers category.

Frequently asked questions

What's the biggest gap between B.Tech graduates and what companies actually need?

The gap isn't in technical knowledge. Most candidates who clear the technical screen have adequate fundamentals. The gap is in the ability to be productive on real code in a real team: navigating a codebase you didn't write, owning something that can actually break, figuring out what to do next without being told. Most programmes build knowledge. They don't build the discipline of working.

Why do most engineering programmes miss the production mindset?

The structure of most programmes doesn't create the conditions for it. Every assignment is isolated: graded by the teacher, then archived. No code runs on real users. No one else depends on it. The feedback loop is the mark, not the failure of the system. Four years of that leaves the accountability muscle unbuilt. A student can know exactly what a linked list is and still freeze when asked to extend one in an unfamiliar codebase under real pressure.

Does starting real work in Year 1 actually make a difference to outcomes?

Yes, and not for the reason most people assume. It isn't the specific stack that matters. A Year 1 stack will be outdated by Year 4. What matters is the habit it builds: shipping something that needs to work, that someone checks, that could break. That discipline, built early and repeated, is what turns a student who knows into a graduate who can do.

How does Kalvium close this gap?

Kalvium students ship working software from Semester 1. The DOJO runs daily coding practice six days a week, building the habit of being uncomfortable and working it out anyway. From Semester 3, students build alongside industry partners on real projects where code is reviewed by someone who isn't awarding marks. The 2026 batch was 82.40% placed as of March 2026, with a median of 16.5 LPA, and companies like Morgan Stanley, PhonePe, Thoughtworks, and Yellow.ai on the recruiter list.

What should a student or parent look for to know if a programme closes this gap?

Ask one question: what does a student ship in Year 1, and what happens when it breaks? If the answer is a classroom project that gets graded and archived, the gap won't close. If the answer is working software with real constraints, reviewed by someone who treats the student as a colleague rather than an examinee, you're looking at a programme that has a real chance of closing it.