Artificial Intelligence and Machine Learning Jobs: What Is Actually Hiring

"Artificial intelligence and machine learning jobs" is not one search, even though it looks like one. It covers a data scientist building forecasting models, an MLOps engineer keeping a training pipeline running, a product manager who has never written code, and an entry-level analyst labelling data for a fraction of any of their salaries. Before you can answer "is this hiring", you need to know which of those you are asking about.

The titles hiding under one search term

Roughly, the postings sort into these groups. They ask for different backgrounds, pay differently, and hire at very different volumes.

• —Data scientist. Builds models to answer a specific business question — who will churn, what will demand look like next quarter. Usually wants statistics, SQL and a portfolio of analyses, not deep learning research.
• —Machine learning / AI engineer. Takes a model into production: serving it, monitoring it, keeping it fast and cheap. Closer to software engineering than to research. Covered in detail in AI engineer jobs, including the four sub-roles that title actually hides.
• —MLOps / ML infrastructure. Pipelines, versioning, GPUs, cost. The plumbing that makes the other two roles possible, and often the best-paid of the group, because it draws from the smaller pool of people who are comfortable with both distributed systems and the model layer on top of them.
• —Data / model evaluation and annotation. Labelling datasets, writing test cases, checking whether a model got better or worse. The easiest entry point, and the most under-titled — postings for it rarely say "AI" at all.
• —AI product and program roles. Deciding what to build, not building it. Wants domain knowledge and the judgement to say no to a bad idea more than it wants code.

A posting titled simply "AI/ML Engineer" without further detail is usually shorthand for one of these, chosen by whoever wrote the req, not a sixth category. Read the responsibilities section before the requirements section — it tells you which of the five you are actually applying to.

What the growth numbers actually say

Two things are true at once here, and it is worth holding both.

The first is that the underlying demand is real and measured. The US Bureau of Labor Statistics projects data scientist employment growing 34% from 2024 to 2034 — from about 245,900 people to 328,300 — with roughly 23,400 openings a year on average over the decade. That makes it one of the fastest-growing occupations the BLS tracks, in either direction, AI included or not.

Indeed Hiring Lab found a similar pattern from the posting side: machine learning engineer job postings sit around 59% above their February 2020 level, even while software development postings overall remain roughly 27.5% below that same baseline. Machine learning specifically is one of the few corners of tech hiring that is not just recovering — it is growing past where it started.

The occupation is genuinely growing. That is a different claim from "there is an easy way in", and the postings answer the second question less kindly than the first.

It also sits next to a real, separate number worth knowing: AI is simultaneously blamed for tens of thousands of layoffs elsewhere in the economy, covered with its own sourcing in AI and job loss. Both are true at once, because the roles growing and the roles being cut are mostly not the same roles.

The entry-level reality

Most of that growth is not landing on people with zero experience. Job postings for data science and ML roles routinely ask for a related degree — computer science, statistics, or a quantitative field — but the harder filter in practice is evidence of work: a project you can point to, not just a transcript. Junior roles specifically titled "machine learning engineer" are a genuinely small slice of the postings; the more common route in is a data analyst or software engineering role that grows into ML responsibilities over a year or two, not a direct hire into the title.

That is a different pattern from a field where a certificate substitutes for a decision, which is worth reading in full if you were weighing one — see do AI certificates mean anything. Here, the credential gets you a conversation. A working project gets you the interview.

What the postings ask for, once you filter by role

Across the data scientist and applied ML postings, a fairly stable order shows up:

1. 01SQL and basic statistics. Not glamorous, checked in nearly every screen, and the thing self-taught candidates most often skip in favour of the model itself.
2. 02One programming language, almost always Python, well enough to read someone else’s code and debug it — not necessarily to design a new architecture.
3. 03A portfolio of finished, explained work. One clean analysis with a clear question, a method, a result and an honest account of its limits, beats five half-finished notebooks.
4. 04Communication. The single most repeated soft requirement on data science postings specifically, because the job is explaining a result to someone who did not build the model.
5. 05Deployment and evaluation, for the engineering-leaning roles. Whether a change made a system measurably better, and whether it runs reliably once shipped.

Notice what usually is not there for the applied roles: deep theoretical grounding in how a transformer or a gradient boosting algorithm works internally. Useful background, rarely the gate. Postings do sometimes ask for it as a screening filter on paper, but interview loops for applied positions consistently spend more time on the evaluation and deployment questions above than on the internals of the model itself.

Building the portfolio that actually gets read

A portfolio project only counts if someone else can look at it and understand what you did and why. AI project ideas that actually teach you something covers the difference between a project that is easy to start and one that is worth finishing — the classifier with an honestly reported accuracy and a discussion of the misclassified cases beats a flashier demo with no evaluation at all, because evaluation is exactly the skill hiring managers say is missing.

The route in, by starting point

• —Already write SQL or code for a living. Closest route: add one or two finished ML projects with clear evaluation, and look at analyst or "ML-adjacent" postings inside your current industry before chasing the pure title.
• —Coming from a quantitative but non-technical background — finance, research, operations. Your domain knowledge is worth more than it feels like; pair it with one project in your own domain’s data rather than a generic tutorial dataset.
• —Starting from zero. The evaluation and annotation roles are the realistic first door, not the engineering ones, and they teach you what "good" looks like before you try to build it — though the paid contract version of that work, covered in AI tutor jobs, is a side income more than a route into a full-time title.
• —Already technical and want to move fastest. MLOps and infrastructure roles hire from general software and platform engineering backgrounds more readily than data science does, because the core skill — reliable systems — transfers directly.

The alternative worth taking seriously

For a large share of people searching this phrase, retitling into a full ML role is not the only way to capture the upside. PwC’s 2026 Global AI Jobs Barometer, built from close to a billion job postings across 24 countries, found roles requiring AI skills carry a 62% wage premium over otherwise comparable roles — up from 57% the year before — with those postings growing 69% against 9% for the wider market. That premium is not confined to engineering titles; it shows up in finance, operations, marketing and administration roles where someone simply uses these tools well.

That is the same conclusion reached from a different angle in AI-proof careers: being the person in your existing field who actually uses AI tools competently is a shorter, more available path to the wage premium than a full career change into a title with a long queue behind it.

The short version

Artificial intelligence and machine learning jobs are a real, measurably growing category, not a hype label — the BLS and the posting data agree on that. But the title covers five different jobs with different bars, the entry-level door is narrower than the headline growth suggests, and a finished, honestly evaluated project consistently beats a certificate at getting you through it. If a full retitle is not the fastest route for you, the wage premium for using these tools well is available from where you already stand.

Coursium teaches the practical use of these tools — the requests, the checking, the workflows — for people building that second path. Stay ahead of AI by learning the tools on your phone.