Who actually needs a degree

The answer changes by role

Research scientist. A PhD is effectively required. Not because the maths is unlearnable elsewhere, but because the hiring signal labs use is published work, and publishing is what a PhD is for. This is the one role where the credential is close to a hard gate.

ML engineer. A degree helps and is not required. What is required is that you can write production code, reason about systems, and show something you shipped. Plenty of people arrive from backend engineering with no ML coursework at all.

Data engineer. The least credential-bound serious role in the field. It is software engineering with databases. Employers test it, and tests do not care where you studied.

Data analyst. Wide open. The gate is SQL and the ability to explain a number to someone who does not want to hear it.

AI product. Bound by experience, not education. Very hard as a first job, very reachable as a second or third.

What a degree actually buys

Three distinct things, and people confuse them constantly.

1. 1Legal eligibility. For a work visa in many countries, a recognised degree is not a preference but a requirement in the rules. A US H-1B, a German or EU Blue Card, several Gulf work permits. If your plan involves relocating, check the actual regulation before you plan around a portfolio.

1. 1Passing an HR filter. Large banks, telecoms, government bodies and consulting firms filter applications on a qualifications field before any human looks. Your project never gets opened. This is not a judgement about your skill; it is a form field.

1. 1Actual knowledge, mostly for research: linear algebra, probability, optimisation, and the habit of reading papers. Real, and learnable outside a university with a great deal more discipline than most people have.

Notice that only the third is about learning.

Certificates and bootcamps, honestly

A cloud certification (AWS, Azure, GCP) has genuine currency in enterprise and consulting hiring in India, the Gulf, Nigeria and parts of Europe, because staffing firms bill clients partly on certified headcount. That is a real, narrow, commercial reason.

Outside that, a certificate mostly tells an employer that you paid for a course. It is not worthless — it is weak. It cannot carry an application by itself, and no volume of them adds up to one thing you built.

Bootcamps vary enormously and their outcome claims are marketing. If you consider one, ask for the placement rate calculated over everyone who enrolled, not everyone who graduated, and ask which employers hired last year's cohort by name.

If you do not have a degree

Aim at the places where the filter is not there: startups, small and mid-size companies, agencies, remote-first firms, and internal moves inside a company that already employs you. Skip the graduate schemes at large corporates; they exist to hire graduates and no amount of GitHub changes that.

And be precise about what happened when you get rejected. "Rejected in forty minutes with no interview" is a filter. "Rejected after a technical round" is feedback. Only the second one is about you.