Everyone wants you to specialize early.
Pick a lane. Go deep. Become the embedded systems person, the ML person, the power electronics person. The logic is straightforward: specialization is legible to employers, it compounds, it gives you a clear trajectory.
I've been resisting this advice for two years. Not out of indecision, but out of a working theory that curiosity, applied at the right time, is worth more than early certainty.
The problem with early specialization
Specialization answers the question: "What do you know?" But engineering problems don't respect disciplinary boundaries. The drone that can't hover isn't a controls problem or a firmware problem or a sensor fusion problem; it's all three, simultaneously.
When you specialize early, you develop fluency in one language. That's valuable. But you also develop a tendency to reach for the same tools regardless of the problem. The controls engineer sees a control problem. The ML engineer sees a data problem. The hardware engineer sees a hardware problem.
The most interesting problems sit at the intersections. They need someone who can see all the constraints at once.
What curiosity actually buys you
Curiosity isn't the same as indecision. It's a disciplined practice of staying early in the question, of resisting the urge to declare "I know this" before you've actually explored the shape of it.
In practice, for me, this has meant: one semester digging deep into circuits and control theory, the next building a computer vision pipeline on an embedded platform. A project on RC vehicle control, then a study group on transformer architectures. Following the thread wherever it leads, long enough to understand the non-obvious constraints.
The result isn't expertise in everything. It's something closer to taste: the ability to recognize which tools are the right ones for a given problem, even when you haven't used them before.
Specialization is downstream of taste
Here's the model I've settled on: specialization is downstream of taste. You need enough breadth to develop taste, to understand what kinds of problems are genuinely hard, which approaches have fundamental limitations, and what well-designed technology feels like in practice. Once you have taste, specialization becomes a deliberate choice rather than an accident of what you happened to study first.
The timing matters. There is a window, probably during university, maybe early career, where staying curious is worth more than the opportunity cost. Before you have taste, specialization just means drilling deeper into the first thing you tried.
I'm still in that window. Trying to make the most of it.