Artificial Intelligence
Courses on machine learning, deep learning, and putting language models to work in everyday practice.
Verified courses first, then the rest of the category
A five-chapter course on building and deploying AI agents, with three bonus units on fine-tuning, observability and a game application. Chapters are designed for about three to four hours a week, and two free certificates are offered with no deadline. Stated prerequisites are basic Python and basic knowledge of language models.
Verification record: TMC-2026-00016Verified Re-check due Direct course review
Twelve lessons on optimising for answer engines rather than for the classic ten blue links, which is one of the newer areas where published training is still thin. Free on the Ahrefs Academy.
Verification record: TMC-2026-00019Verified Re-check due Direct course review
A course on applying transformer models to audio data, taught through the Hugging Face libraries. Free and openly readable, covering a domain most general machine learning courses leave out.
Verification record: TMC-2026-00017Verified Re-check due Direct course review
Harvard's AI course, published free as OpenCourseWare. Seven lecture topics work through search, knowledge representation, uncertainty, optimisation, machine learning, neural networks and language, with Python projects attached to each. It is explicitly a follow-on rather than a first course: the stated prerequisite is CS50x or roughly a year of Python experience, and the material assumes the learner can already write and debug their own programs.
Verification record: TMC-2026-00006Verified Re-check due Direct course review
Hugging Face’s twelve-chapter course on large language models, taught through its own Transformers, Datasets and Tokenizers libraries. It moves from how transformer models work to fine-tuning, dataset curation and reasoning models, and states a workload of roughly six to eight hours per chapter. Free with no ads and no certification currently offered.
Verification record: TMC-2026-00015Verified Re-check due Direct course review
A lecture series that builds neural networks from nothing, in code, one line at a time. It opens by writing an automatic differentiation engine by hand and ends by building a GPT and then its tokeniser, with the intermediate lectures constructing a character level language model in progressively more capable forms. The videos are published free on YouTube with accompanying code. The stated prerequisites are solid Python and introductory calculus, and the format is long form: single lectures frequently run over two hours.
Verification record: TMC-2026-00007Verified Re-check due Direct course review
A free deep learning course that starts by training and deploying a working model in the first lesson and only then works backwards into the theory. It covers computer vision, natural language processing, tabular data with random forests, collaborative filtering and building a model from scratch, taught in PyTorch and the fastai library through Jupyter notebooks. It assumes about a year of coding experience, preferably Python, and no mathematics beyond school level. The edition published at the time of checking is labelled 2022.
Verification record: TMC-2026-00014Verified Re-check due Direct course review
A course on post-training AI models, published free by Hugging Face. Narrow by design: it addresses what happens after a base model exists rather than how to train one from scratch.
Not independently verified
A twenty-five minute course on Asana’s AI Teammates feature, covering how the tool delegates recurring work. Product-specific rather than a general AI course, and free with a free Academy account.
Not independently verified
A short Google course on using AI tooling in a job search, which is a genuinely current subject with little published training.
Not independently verified
Close to ten hours on agentic AI systems, pitched at intermediate level, covering the design of systems that plan and act rather than only respond. Published by DeepLearning.AI.
Not independently verified
Vanderbilt on agentic AI for people deciding whether to deploy it, rather than for the engineers building it. Delivered through Coursera; the browse listing publishes no price.
Not independently verified
A short Google specialization on practical AI use at work, sitting below its full AI certificate and intended as an orientation rather than a qualification.
Not independently verified
A five-hour introduction to working competently with AI, published by a university rather than by a vendor with a product to sell. Free from The Open University, startable without signing up, with a free statement of participation on completion.
Not independently verified
Wharton’s specialization on applying AI in a business context, addressed to managers deciding what to do with it rather than to the engineers who build it.
Not independently verified
Andrew Ng’s non-technical introduction to AI, written for people deciding what AI means for an organisation rather than for people building it.
Not independently verified
Applying AI tooling to search work, one of the newer subjects where published training from an established SEO company is still scarce. Included in the Yoast premium plan, which the academy prices at 118.80 US dollars per year for every course plus the company plugins.
Not independently verified
A badge path on applying AI features to everyday work inside Asana. Free with a free account, and product-focused rather than a general introduction to AI.
Not independently verified
Setting limits on how AI tooling is used across a team, which is a governance question rather than a technical one. Free from Webflow University, with an account offered to track progress.
Not independently verified
An hour on using AI in teaching, addressed to educators rather than to students, covering classroom applications and the judgement calls they involve. From Canva Design School.
Not independently verified
A twenty-seven minute crash course on Semrush’s AI PR toolkit, taught by Mariah Magazine across nine lessons. Free with a Semrush account, and product-focused rather than general PR training.
Not independently verified
A seven-hour beginner course on prompting, aimed at general users rather than developers, which is a narrower and more useful framing than most introductions to the subject. Published by DeepLearning.AI.
Not independently verified
Eleven and a half hours teaching Python specifically through AI use cases, rather than teaching Python first and AI later. Beginner level, from DeepLearning.AI.
Not independently verified
Nine lessons on adapting search marketing practice to AI-mediated search, taught by Leigh McKenzie. Short at around forty minutes, and free with a Semrush account.
Not independently verified
About this subject
Courses on artificial intelligence, spanning three quite different things that share a name: the mathematical and algorithmic foundations, the engineering of building and deploying models, and practical training on using AI tools competently at work. The subject also moves faster than anything else in this directory, which changes how you should read a course listing.
Guidance
Verification tells you that a course delivers the material it advertises. It cannot tell you whether the tools it teaches will still be current when you finish. These are the things worth checking yourself.
A course on how transformers work and a course on how to prompt a particular assistant are both legitimate and almost unrelated. Fundamentals stay useful when products change and cost more effort up front. Tool training is immediately applicable and has a shorter shelf life. Neither is a substitute, and the biggest source of disappointment in this subject is buying one while wanting the other.
This is the fastest-moving subject in the catalogue. A course recorded two years ago may still be excellent on backpropagation and useless on which model to use. Where a provider publishes a last-updated date we record it, and the summary usually says whether the course is taught around a specific product or around concepts that outlive it.
Much of the best AI training is published by the companies that make the models and libraries, and it is often genuinely the primary source. It is also, by construction, about their ecosystem. That is not a reason to avoid it, but it is worth knowing whether you are learning a field or a product. Each listing names the publisher, and where a course is built around one vendor the summary says so.
Courses in this subject range from requiring no technical background at all to assuming linear algebra, calculus and a year of Python. Providers usually state prerequisites plainly, and they mean them. The level on each listing reflects what the provider states.
AI attracts marketing about replacing work, multiplying output and career transformation. Treat those as claims about the future rather than descriptions of the course. What a course can honestly offer is that you will understand something or be able to do something specific, which is the part worth reading closely.
Standards
The published standard applies unchanged. These are the checks that tend to matter most in this subject.
A verification is dated, and in this subject the date matters more than in any other. It says nothing about whether the models, tools or techniques taught are still the ones you should be using.
Questions
Yes, and several in this category are explicitly written for a general or business audience. The level and the summary on each listing say who the course is pitched at, and courses assuming Python or mathematics normally state that prominently.
It depends entirely on what the course teaches. Foundations age slowly. Anything tied to a specific model version, product interface or pricing tier can be out of date within months. The listing names the publisher and, where stated, the last update.
Not at all. They are often the most accurate source on their own tools. Just read them as what they are: excellent on that ecosystem, and not a survey of the field.
Related