There’s a stretch in most young-driver programmes, F1 or otherwise, where a promising 19-year-old has to make a choice nobody else can make for them. Stay in the simulator, where the stakes are contained and the feedback comes from your own team. Take a loan drive at a smaller team, where a worse car forces you to extract more from less. Or skip straight to building your own team from scratch, where nobody’s grading you because there’s no syllabus, just the results.
None of those paths is the “correct” one. Different drivers need different things at different points. A driver who’s never actually raced under pressure needs the loan drive more than another lap in the simulator. A driver who’s raced plenty but never had to make a strategy call themselves needs the opposite.
I think about that structure whenever I look at what happens to Kalvium students in Semester 3. From that point on, the schedule shows a new subject, listed plainly as Internship or Project. Underneath that one line sit four genuinely different tracks: Simulated Work, Internship, Open Source Projects, and Build AI-native products. They don’t rank by difficulty. They train different things.
Four different rooms, four different judges
Here’s the distinction that actually matters more than any surface-level difference between the four tracks. Who’s evaluating the work. And how much of the decision is the student’s own.
Simulated Work puts students in in-classroom sprints built to replicate a company engineering environment. There’s a brief, a team, a delivery cadence. But the reviewer at the end of the sprint is a Kalvium mentor, someone who’s invested in the student getting better, not someone deciding whether to renew a contract. The feedback loop is real, but it’s contained. Mistakes here cost a sprint retro, not a client relationship.
Internship hands that same kind of work to a company that owes the student nothing. The brief still comes from someone else, mostly, but the person reading the pull request or reviewing the sprint isn’t grading for growth. They’re evaluating whether the work is useful to them. That’s a different kind of pressure, even when the technical task looks similar to Simulated Work on paper.
Open Source Projects goes further. The codebase belongs to strangers. The student has to read an architecture they didn’t design and work out where a contribution actually fits. Then they have to defend that change in a pull request review conducted by people who’ve never met them, and have no institutional reason to be patient. There’s no mentor in the loop softening the landing. The review either holds up or it doesn’t.
Build AI-native products, the entrepreneurship track, removes the last piece of scaffolding. In the other three tracks, someone else has already decided what needs to be built. Here, the student decides that too. It’s not just execution under scrutiny anymore. It’s ownership of the call that gets scrutinised.
Why this is a training question, not a ranking
Most 19-year-olds facing this choice ask the wrong question. I’d guess it’s usually some version of “which one’s hardest” or “which one looks best on a resume.” I say that as someone who’s asked plenty of wrong questions myself at that age.
The better question is what a student hasn’t been forced to do yet.
A student who’s spent two semesters getting instant, precise feedback from DOJO and LiveBooks has been training in a specific kind of environment. A problem either passes or it doesn’t. That’s a very short, very clean feedback loop. That’s valuable. It’s also incomplete. Simulated Work extends that loop slightly, into a team setting, while keeping the judge friendly. For a student who’s technically sharp but has never had to defend a decision to someone outside their own head, that’s the right next room.
A student who’s comfortable being evaluated, who’s already had mentors poke holes in their thinking for a year, might get more out of a different room. Internship or Open Source, where the evaluator isn’t rooting for them. The absence of a friendly reviewer is the whole point. It’s the first place a student finds out whether their work holds up when nobody’s grading on effort.
And Build AI-native is for a different gap entirely. It’s not really about coding skill at all. It’s about whether a student can sit with the discomfort of “I decided this, and it might be wrong,” which is a muscle that execution-only tracks never touch, no matter how technically demanding the execution gets.
The part I’d be honest about
Here’s what I can tell you with confidence from what’s documented: what each track involves, and roughly what kind of pressure each one applies. What I can’t tell you, because it isn’t published anywhere I’ve seen, is exactly how the selection process works. Whether it’s a student’s free choice, a mentor recommendation, some mix of the two, whether a student can switch tracks mid-semester if the fit is wrong. Those are real questions a family should ask the admissions team directly, not questions I want to guess an answer to here.
What I’ll say instead is what I actually believe about the four tracks as a set. The fact that Kalvium runs four of them, rather than one default path, is itself the more interesting design decision. Most programmes that offer “real-world exposure” mean one flavour of it, usually an internship, and call it a day. Structuring four distinct kinds of scrutiny, mentor-graded, company-graded, stranger-graded, self-graded, is an admission that different students arrive at Semester 3 with different gaps. A programme that only offers the internship track is implicitly betting every student needs the same thing next. That bet is usually wrong.
What doesn’t change underneath the tracks
The track a student picks changes who’s judging the work. It doesn’t change everything else running underneath it. DOJO keeps running, same belt structure, same daily practice, because the technical foundation underneath all four tracks still needs maintaining. The squad structure keeps running too, though what the squad’s actually discussing changes once the problems stop having one correct answer. And the Internship or Project module itself doesn’t end at Semester 3. It runs every semester from here through graduation, so whichever track a student starts in isn’t a single bet locked in for four years. It’s the first of several.
The Semester 3 curriculum and the mechanics of that first shift are covered in more detail in what Semester 3 at Kalvium actually involves. What I wanted to add here is the question underneath the four names: not which track is best, but which room a specific student needs to be standing in next.
I still think about the simulator-versus-loan-drive choice sometimes, mostly because there’s no clean answer to it either. The driver who stays too long in the simulator never finds out what a real strategy call under pressure actually costs. The one who skips it too early sometimes finds out the hard way, in front of people who won’t be gentle about it. Somewhere between those two failure modes is the right call for a specific person at a specific point, and I don’t think there’s a shortcut to figuring out which one you are.
Deepak is a co-founder of Kalvium. He writes at the intersection of education design, philosophy, and how engineering careers are built. Read more from Deepak or explore the B.Tech category.