← All insights

Careers

A realistic first-five-years map of a VLSI career

Careers in chip design are described to students as a single destination. They are actually a set of quite different jobs that happen to share an industry, and the choice between them is usually made by accident — by whichever team had an opening — and then becomes hard to reverse.

The roles that actually exist

Broadly, the work divides into a handful of families:

  • Digital design — turning a specification into synthesisable RTL, and owning its microarchitecture.
  • Design verification — establishing that the design does what it is supposed to, and finding where it does not.
  • Physical design — floorplanning, placement, routing, timing and power closure; turning a netlist into something manufacturable.
  • Analog and mixed-signal design — circuits where the interesting behaviour is continuous, plus the layout craft that goes with it.
  • Architecture and modelling — deciding what to build and proving it makes sense before anyone writes RTL.
  • DFT, validation, EDA and firmware — the adjacent specialisations that projects cannot ship without.

These differ not just in subject matter but in daily texture. Physical design is tool-heavy and iterative. Verification is investigative and communication-heavy. Analog design is slow, deep and intuition-driven. Architecture is argumentative and largely about trade-offs. People who are miserable in one are often thriving in another.

Years one and two: earn credibility

The first two years are spent becoming someone whose work does not need to be checked. Concretely: learning one flow properly, learning the team's tools and conventions, and delivering small things reliably. Nobody expects architectural insight from you. They expect that when you say a block is done, it is done.

The most common early mistake is chasing visibility — volunteering for a high-profile piece before being able to execute an ordinary one. The second most common is passivity: waiting to be taught, when the actual expectation is that you will read the existing code, the previous project's documentation and the specification without being asked.

Year three: the branch point nobody announces

Somewhere around the third year, a quiet fork appears. You will either continue deepening in one domain, or you will start moving toward the interfaces — the parts of the job that involve deciding rather than implementing.

Both paths are legitimate. But domain switching gets meaningfully harder after this point, because you become expensive as a specialist and cheap as a beginner elsewhere. If you suspect you are in the wrong family — and a fair number of engineers do — the third year is roughly the last inexpensive moment to move.

Nobody tells you the switching window is closing. It just gets quietly more expensive each year.

Years four and five: own something

The step that separates a mid-level engineer from a senior one is ownership: being the person accountable for a block, a subsystem or a methodology, including its problems. This is less about technical depth than it looks. It requires being able to estimate honestly, to flag risk early, to say when something will not be ready, and to make decisions with incomplete information.

Engineers who stall at this stage are usually not lacking knowledge. They are avoiding accountability — preferring well-defined tasks handed down, because tasks cannot be blamed on you the way decisions can.

The decisions that matter more than they appear

The first team, more than the first company

A strong team at an ordinary company teaches you more than a weak team at a famous one. In the first few years you learn almost entirely from the people immediately around you.

Choosing depth before comfort

There is always an option that is easier and an option that will teach you more. In the early years, the second compounds. Later on, it gets harder to choose.

Writing things down

Engineers who document, present and explain get pulled into decisions. Engineers who only execute get given execution. It is not fair, but it is extremely consistent.

What this means if you are still studying

Do not try to pick your family at nineteen on incomplete information. Do build the common foundation — digital design, computer architecture, one programming language you are genuinely fluent in, and comfort with Linux and scripting. Then get close enough to real work — through a project, an internship or a serious course — to find out which of these jobs you actually like doing on a Tuesday afternoon.

That last question is the one that determines a career, and it is almost never asked.


Our programmes are built around exactly this problem — see Semicon Gurukul, or download the VLSI entry roadmap.