Do you need maths to become an AI engineer?
The Ship AI Engineer team · Published 2026-08-07
Most people who want to build with AI never write the first line, and it is almost never because the work turned out to be too hard. They stop earlier, at a list of prerequisites somebody else wrote: one that describes a neighbouring job. Here are the six questions that actually stop people, answered as directly as they deserve.
Do you need to be good at maths?
No. You need arithmetic and a tolerance for spreadsheets, and that is close to the whole of it.
The maths on every roadmap (linear algebra, calculus, probability) is what you need to build a model: to design an architecture, to understand why a training run diverges, to read the paper it came from. If you are calling a model that already exists, none of that appears in your day. What does appear is multiplication: tokens per request times requests per day times a rate, which decides whether your design survives contact with a real month. That is arithmetic, and you can do it in your head or in a browser.
There is one honest caveat. When you start measuring whether your system works, you will meet a handful of ideas that look statistical: precision and recall, and why an average can hide a failure. They are worth an afternoon, not a semester, and they are much easier to learn when you already have a broken system in front of you asking the question.
Do you need a computer science degree?
No, and this is more clearly true here than in most of software.
The reason is uncomfortable but useful: the field is young enough that almost nobody senior in it studied it. The people building with language models today learned in the last two or three years, from documentation and from things they broke, because there was nothing else to learn from. A degree still teaches things worth having (how a computer actually works, how to reason about algorithms) but it is not a gate here, and nobody is checking.
What is checked is whether you can show something that runs. A repository someone can clone, read and start is the credential that gets looked at.
Do you need to learn machine learning first?
No, and starting there is the most common way people lose two months.
Machine learning is the discipline of building models from data. AI engineering is the discipline of building products from models. They share vocabulary and almost no daily work. You can spend a year on gradient descent and be no closer to shipping something that answers questions about your own documents, because the hard parts of that (getting real files into clean text, deciding what to retrieve, checking the answer is grounded) are not machine learning problems.
If you later want to train models, the ML path is there and it is a real career. It is a different one, and doing it first is not a prerequisite. It is a detour.
How much Python do you need?
Enough to read code you did not write, and to fix it when it breaks.
That is a lower bar than it sounds and a different bar than the usual one. You need variables, lists and dictionaries, functions, loops, and the ability to read a traceback from the bottom up. You do not need decorators, metaclasses, asynchronous generators or the type system.
The skill that has quietly become essential is the one no course had a name for five years ago: reading code a model wrote for you and being able to tell whether it does what it claims. A model produces plausible code, and plausible parts company with correct exactly where you stop looking. That is a reading skill, and it is why "just get an AI to write it" does not remove the need to learn.
Do you need an expensive computer or a GPU?
No. The laptop you have is almost certainly enough.
Training models needs GPUs; calling them does not. When you call a hosted model the computation happens somewhere else and your machine sends an HTTP request. If you want to run a model locally (which is a genuinely good idea for learning, because it costs nothing per request and no data leaves your desk) a recent laptop runs a small one, slowly and well enough to learn on.
The one place hardware becomes real is fine-tuning, and by the time that is the right answer for you, renting a machine by the hour is the normal way to do it.
Do you need to spend money on API credits to learn?
Not to start, and less than people expect afterwards.
A model running on your own machine costs nothing per request and is enough to learn every idea in the first half of this field: prompting, structured output, embeddings, retrieval, tool calls. When you do move to a hosted model, the thing that actually decides your bill is not the price per token but how many tokens you send, how often you send the same one twice, and whether anything stops a loop you did not intend. Those are design decisions, and they are learnable before you spend anything.
You can try that decision right now, with the arithmetic switched on and nothing to sign up for, in the cost game.
Where to start instead
If the prerequisite lists put you off, start from the shape of the field rather than from a syllabus: the knowledge map is 105 ideas, each defined in a sentence, free to read. And if you want to know why the ranked roadmaps disagree with this page, that has its own answer.
Keep reading
2026-10-06
Resources worth following as an AI engineer: what to read, and when
A new post, model or tool appears in this field every day, and trying to follow all of it is the fastest way to learn none of it. A handful of resources that genuinely help, and the stage of learning at which each one is worth reading.
2026-10-05
Being an engineer in the age of AI: code got cheap, structure didn't
Writing code has never been this fast. But code produced quickly amplifies whatever structure sits underneath it. Why engineering matters more today, not less, and where to go deeper.
2026-10-02
Shipping an AI application: spending caps, caching, monitoring and model changes
In an ordinary web app every request costs roughly the same. In an app with a language model inside, it doesn't. Capping cost when you ship, using caching safely, watching the right things, and being ready when the provider retires a model.
2026-09-30
An AI engineer portfolio: what a project that gets you hired looks like
A chatbot demo impresses nobody anymore. Six signs that make an AI project credible, five project ideas that each show a different skill, and a README someone can read in two minutes.
2026-09-27
Prompt injection and abuse: why the server has to be the single source of truth in an AI app
An application with a language model inside carries every security problem of an ordinary web app and adds new ones. Why prompt injection can't be fully prevented, how to limit the damage, protecting the endpoint, and why the model's output can't be trusted.
2026-09-23
Building your first test set for an LLM application: a guide to evaluation
You changed the prompt and the answers look better. Do they? Building a first test set, taking a baseline, seeing what the average hides, the limits of one model grading another, and watching the system once it's live.
2026-09-19
When should you fine-tune, and when is it a waste of time?
Training a model on your own data is an attractive idea. Most of the time it's the wrong tool. What fine-tuning teaches and what it doesn't, how it compares with prompting and RAG, data quality, leakage, LoRA, and licences on open models.
2026-09-14
7 questions to ask before you pick an AI engineering course
Seven questions worth asking before you spend money and months on an AI engineering course. A short answer to each, and the concrete sign you can check in the syllabus before you buy.
2026-09-09
AI with image and audio input: the costs and failures your tests don't show
Handing a model a photo or a recording isn't the same as sending text. Request size, unreadable images, sentences invented in silence, generation that takes seconds, and an image paid for again on every turn.
2026-09-04
When don't you need an agent? Building and stopping a tool-using system safely
"Agent" is today's most requested and most easily misbuilt thing. The difference between a fixed workflow and an agent, why tool descriptions are critical, how to stop the loop, and how to protect irreversible actions.
2026-08-31
Why does RAG give wrong answers? Most failures are in retrieval, not the model
When a system that answers from your documents says something wrong, the model is the first suspect. Most of the time the culprit is the chunk it was given. Finding where the failure is, containing invention, citing sources and measuring properly.
2026-08-26
How to become an AI engineer: a roadmap that builds products, not models
Most roadmaps in the search results describe a different job. What the AI engineers being hired today actually know, the order to learn it in, and what makes each stage "done".
2026-08-21
When do you actually need a vector database? A practical guide to embeddings
The first step people take after deciding to build semantic search is usually picking a vector database. Most of the time it's an early step. What an embedding is, what a similarity score doesn't tell you, and the real reasons to move to a database.
2026-08-17
Managing prompts like code: versions, templates, schemas and tests
A prompt that works in a chat window quietly breaks inside an application. Keeping prompts in the repository, separating user input from instructions, validating the answer against a schema, and shipping a change only after measuring it.
2026-08-12
Why LLM APIs break under real traffic: rate limits, retries and timeouts
Three errors you never saw in development arrive together in the first busy hour. Telling 429 from 5xx, retrying properly, setting timeouts, and keeping the product up when the provider goes down.
2026-08-07
An AI engineer is not a machine learning engineer
Search "how to become an AI engineer" and the first page teaches linear algebra, CNNs and training loops. That is a different job. Here is what the work actually is, and what you can skip.