Yes — and in 2026 it is the normal route, not the exception. In our count of 315 AI job ads from July to September 2026, a degree of any kind was mentioned in 12% of postings; a portfolio, GitHub profile or open-source work in 17%; and production or shipped in 39%. Employers write down what they screen on. The one who wants a computer science degree says so; the other seven in eight are asking to see what you have built.
That is the market-wide picture. The Australian version has two wrinkles worth knowing before you plan around it.
Do Australian employers care more about degrees than employers elsewhere?
Slightly, in three places: government and defence, the big four banks' graduate programmes, and any role sponsoring a visa. Federal and state departments often list a degree as a "desirable" rather than "essential" criterion, but their recruitment panels score against the criteria as written. Graduate programmes at the banks and the big consultancies are degree-gated by design — that is what "graduate" means there. And a skilled visa (subclass 482 or 186) typically needs a qualification or a formal skills assessment, which is a legal requirement, not an employer's preference.
Outside those three, Australian tech hiring runs on the same signals as the US and UK: what you shipped, whether you can explain it, and whether a referee will vouch for you. Sydney and Melbourne startups, the mid-size SaaS companies (Atlassian, Canva, SafetyCulture, Culture Amp and their peers) and the consultancies' delivery teams hire non-graduates routinely for engineering roles. They just do not advertise it.
What does an AI engineer actually do, and what will I be tested on?
Software engineering with a model in the loop — not model research. The most common titles in our sample were software engineer at an AI company (62 ads) and AI engineer (48), against 13 for ML engineer and 2 for research scientist. The skills those ads name most are Python (38%), TypeScript (19%), agents and tool use (39%), evals (23%), RAG (11%), AWS (16%), Docker (10%) and Postgres (10%). PyTorch appears in 5%.
Interviews for these roles have converged on a shape: a take-home or live build of a small model-backed feature, a conversation about how you would know it works, and a system-design discussion about cost, latency and failure. Nobody asks you to derive backpropagation. The AI Engineer role page lists the day-to-day and the curriculum modules that map to it.
How long does it realistically take without a degree?
Six to twelve months of focused building if you can already program in one language; twelve to eighteen if you are starting from zero. The people who make it in six months are almost always career-adjacent — a data analyst, a QA engineer, a sysadmin, a front-end developer — and they spend that time building three or four deployed projects rather than collecting courses. The people who take two years are usually the ones who spent the first year watching videos.
A structured twelve-week bootcamp at 15 hours a week is 180 hours. That is enough for six deployed projects and a portfolio you can interview on; it is not enough on its own for someone who has never written a loop. Be honest with yourself about which you are — the free skill checks exist for exactly that purpose.
What should my portfolio contain?
Three deployed things, not thirty repositories. A hiring manager opens the first link and gives it ninety seconds. Make those ninety seconds count:
- A model-backed application with a live URL — a real user can type into it, and it does something useful. A support-ticket classifier for a real (or realistic) dataset, an agent that books something against a real calendar, a retrieval system over a corpus you care about.
- An eval set and a number. "It answers 84% of the 200 test questions correctly, and here are the 32 it gets wrong and why." That one sentence does more than any certificate, because most candidates cannot produce it.
- Something that broke in production and what you did. A short write-up. Employers hire for the second incident, not the first.
The portfolio piece goes into the format. Every Square 1 bootcamp is built as six of these projects, each deployed, each graded against a rubric you can show.
Will I be paid less without a degree?
Not measurably once you are in. Australian tech salaries are set by role, level and city far more than by credential, and after the first job nobody asks. Where the degree costs you is the first job — a narrower set of doors, and a higher bar of proof at the ones that are open. Expect to enter at a junior or mid level on the strength of your portfolio, then move at the same speed as everyone else.
Skillsoft's 2025 IT Skills and Salary report has a third of IT decision-makers valuing a certified team member at US$25,000 or more a year — worth knowing if you are weighing a vendor certification as a door-opener. The certification piece covers which ones Australian ads actually name.
Should I do a degree anyway?
Only if you want one of the three doors above, or you want the maths. A Graduate Certificate in AI or data science from an Australian university is a real credential that takes six months part-time and opens the government and graduate-programme doors without committing you to a full masters. If your goal is a startup or a SaaS engineering team, spend the same money and time on building and you will get there faster.
The difference between an AI engineer, an ML engineer and a data scientist matters here: the data-science and research-scientist paths still lean on degrees. The AI-engineering path does not.
Where do I start this week?
Take the free AI Engineer skill check, then build the smallest possible deployed thing — a single endpoint that calls a model, with a test set of twenty questions — and put the link on your LinkedIn. That project is worth more than the next month of reading. The hiring board lists roles from employers who have agreed to read graded records instead of CVs; the leaderboard is where your graded work becomes visible to them.