We earn a commission if you buy through links on this page. How we're funded.
Quick answer
AI careers fall into four groups. Engineering roles build with models: AI, machine learning, generative AI, agent, LLM, computer vision, NLP and MLOps engineers. Data roles analyse and model data, and data analysis is among the most accessible entry points into data work. A third group directs or governs AI: product managers, consultants and governance specialists, plus solutions architects, a senior role usually reached after years of engineering, cloud or consulting delivery. Research is the fourth, and its scientist roles usually need a PhD. Choose by where you start: code, data, or experience in a field.
“AI career” covers jobs that have little in common: some are software engineering with a model inside, some are data work, some are product and policy decisions about AI, and one is research. This page maps the 16 roles we have written a full guide for, says in one line what each job is, and points to the guide that sets out the route in.
The AI roles at a glance
The table below lists the 16 roles we have a full guide for, the group we place each in, and what each guide says the job is or asks of you.
| Role | Group | What the job is or asks for | Enrol |
|---|---|---|---|
| AI engineer | Engineering | Builds software that puts machine-learning and LLM capabilities into production; closer to software engineering than to research. | 365 Data Science → |
| Machine learning engineer | Engineering | Builds, trains and ships models that run reliably in production, on real mathematics as well as strong software engineering. | |
| Generative AI engineer | Engineering | Builds reliable applications on foundation models: prompting, retrieval-augmented generation, tool calling, evaluation and deployment. | |
| AI agent engineer | Engineering | A software engineer who ships LLM systems that take actions: tool calling, orchestration, evaluation and guardrails. | 365 Data Science → |
| LLM engineer | Engineering | Works on the model layer itself: data curation, fine-tuning, quantization, inference serving and evaluation. | |
| Computer vision engineer | Engineering | Deep learning for images (detection, segmentation, vision transformers), plus deployment skills such as edge inference. | |
| NLP engineer | Engineering | Language tasks: tokenization, embeddings, transformer models, fine-tuning and evaluation. | |
| MLOps engineer | Engineering | The machine-learning lifecycle as infrastructure: reproducible pipelines, model registries, automated deployment and drift monitoring. | |
| Data analyst | Data | SQL, spreadsheets, one visualization tool and basic statistics, proved with projects that answer business questions. | |
| Data scientist | Data | SQL, Python and statistics applied end to end to messy, real data. | |
| AI product manager | Directing and governing AI | Classic product skills plus enough AI fluency to judge what is feasible, what data is needed and where models fail. | |
| AI solutions architect | Directing and governing AI | Cloud architecture plus practical AI system knowledge, turning business problems into designs; a senior role, rarely a first job. | |
| AI consultant | Directing and governing AI | Domain expertise clients pay for, AI fluency, and the delivery skill to scope a measurable engagement. | |
| AI governance specialist | Directing and governing AI | AI risk frameworks and regulation, with enough technical literacy to interrogate a real system. | |
| Prompt engineering | A skill across roles | Best pursued as a skill inside a larger role (specification, evaluation and workflow design) rather than as a job title. | Coursera → |
| AI researcher | Research | Deep mathematical and engineering skill plus a public record of original work. |
The four groups, and what each one asks of you
Engineering: building with models
Eight of the roles are software engineering with a model inside. They differ in which layer they work on, from applications built on foundation models (generative AI and agent engineers) down to training, serving and operating models (machine learning, LLM and MLOps engineers). Our generative AI engineer guide says most hires come from backend, data or machine learning engineering rather than from research, and that shipped projects matter far more than certificates.
Data: analysing and modelling
Data analysis is the usual first step into data work: SQL, spreadsheets, a visualization tool and basic statistics, proved with projects. Data science adds Python and statistics applied end to end, and our data analyst guide notes that many data scientists began as analysts and moved across after two or three years.
Directing and governing AI
These roles are bought on judgment and field experience more than on code, with one exception. Our governance guide says most people enter from law, privacy, audit, risk or compliance; the AI product manager guide says the role most commonly comes from product management plus AI fluency; and consultants sell domain expertise clients already value, which is why our guide to AI consultant certifications starts with no-code courses. The solutions architect is the exception: its guide expects coding ability at the level of a competent senior engineer and says most architects come from senior engineering, cloud or consulting roles.
A skill across roles
Prompt engineering is in the table because people search for it as a job, but its guide says to pursue it as a skill inside a larger role, such as an AI-fluent marketer, analyst, support lead or engineer, rather than as a job title.
Research
A PhD is the standard route to research scientist roles. Our AI researcher guide says research engineer positions are open to strong engineers with published or reproducible work instead.
Not sure this is the right one for you?
Tell the picker about your background and what you want the certificate to do, and it narrows the list to the one or two courses we would start with. It suggests only our affiliate partners’ courses, and says so before it suggests anything.
Try the AI Certification Picker →Where to start from your background
- You already write software. The engineering roles are the shortest move. Our AI engineer guide puts it at roughly six to twelve months of focused upskilling from working software engineering, and the agent and generative AI roles build on the same base.
- You work with data. Data science and machine learning engineering are the natural next steps: our data analyst guide notes that many data scientists began as analysts and moved across after two or three years, and the machine learning engineer guide gives about a year of focused study from data science. For AI engineering, the AI engineer guide gives nine to eighteen months from a data or analytics role.
- You do not code. Product management, consulting and governance reward experience in a field more than programming; our AI product manager guide says the role does not require code. Solutions architecture is the exception in that group: its guide expects coding ability at the level of a competent senior engineer. Data analysis needs no degree and is among the most accessible entry points into data work, but it does need SQL, the skill its interviews test most. Our list of AI jobs that need no coding goes further.
- You are starting from zero. Our AI engineer and machine learning engineer guides both budget two years or more, because you need to become a competent programmer first (the agent engineer guide says one to two years); our entry-level AI jobs guide covers the first job, and the AI certification roadmap the order to learn in.
- You want research. A PhD is the standard route to research scientist roles; research engineer roles are open to strong engineers with published or reproducible work.
If you are deciding between three of these titles, AI engineer vs ML engineer vs data scientist compares them directly, and the best AI certifications for a career change covers the credentials that help at each step. For pay, see the highest-paying AI jobs.
Frequently asked questions
What is the easiest AI career to get into?
Of the roles here, data analysis is the most accessible technical route in, as our entry-level AI jobs guide puts it, though that guide says most people without a technical background get in fastest through an AI-augmented version of a job they can already do. Our data analyst guide sets out a route that needs no degree: SQL, spreadsheets, one visualization tool and basic statistics, proved with two or three projects that answer real business questions. Among the AI-specific roles, product management, consulting and governance reward experience in a field more than coding.
Do you need a degree for an AI career?
Not for most of these roles. Our data science guide says a degree helps but is not required, and our data analyst guide says most employers screen for demonstrable skills and a portfolio. Degrees still count for some visa-sponsored and regulated roles, and some large employers filter on them automatically. Research is the exception: a PhD is the standard route to research scientist roles and is effectively required at most industrial labs, although research engineer positions are open to strong engineers with published or reproducible work.
How long does it take to move into an AI role?
It depends on where you start. Our AI engineer guide puts it at roughly six to twelve months of focused upskilling from working software engineering, nine to eighteen from a data or analytics role, and two years or more from zero, because you need to become a competent programmer first. From an existing product management role, the AI product manager guide gives six to twelve months; the governance guide gives six to eighteen months depending on your starting point, and the data analyst guide three to six months of part-time study for a first analyst role, plus one to three months of applying.
Which AI career path should I choose?
Start from what you already do. If you write code, the engineering roles are the shortest move; if you work with data, data science or machine learning engineering; if you bring expertise in a field such as law, risk, product or a client industry, product management, consulting or governance. Aim at research scientist roles only if you are prepared for a PhD; research engineer roles are open to strong engineers with published or reproducible work.