Learn what a large language model really does, from tokens to post-training, what it can and cannot do, and how to choose one for your own task with evidence instead of a leaderboard.
Read every lesson as a guest, or sign in to enroll and track your progress.
Enroll to save the lessons that you complete. It is free.
By the end of this course, you will be able to:
A large language model writes fluent text about almost anything. That makes it easy to trust and hard to judge. Two models can give answers that sound equally good, while one of them is wrong twice as often on your own work, costs four times as much, or keeps your customers waiting.
In this course, you learn how a language model works at the level that a person who chooses one needs: tokens, probabilities and the stages of training, without heavy mathematics. You see real models make real mistakes, and you learn which mistakes more context can fix and which need a check outside the model.
Then you compare models the way a careful team does. Larkfield’s help desk is about to answer customers in English and French, and it must choose a model. You build a small test set, score the answers with a rubric before you know which model wrote them, and measure time and cost. The course uses real recorded outputs of hosted models and of open-weight models run on a small desktop computer, so you can do every comparison without an API key.
Finally, you turn the evidence into a decision that someone else can check, with a fallback and the conditions to look again. Diagrams and interactive charts carry the main ideas, so the text stays short.
You completed Essential Mathematics and Statistics for AI, or you can read a probability distribution, a mean and a percentile. Deep Learning and Transformers Explained is useful background but not required. You do not need an API key; every comparison uses recorded outputs, and live calls are optional.
Courses to take first: Essential Mathematics and Statistics for AI.
People who know the basics of probability and statistics, as in Essential Mathematics and Statistics for AI, and must decide which language model a product should use. No programming is needed for the main path; optional notebooks let you run some of the examples.
You can study the ideas of this course without writing code. Read modules 1–4 (about 4–6 hours), answer their knowledge checks, and get a free concepts certificate. You can come back for the practice later.
The paths for people who lead or use AI take this course this way. See the paths
Experienced AI practitioners designed this course. AI helped us draft the content; people chose the teaching approach and the examples, and reviewed every lesson. The outputs that you see are the actual results of running the code. How we make our courses
See what happens inside a language model at the level you need to use one well. Text becomes tokens, the model gives each possible next token a probability, and training stages shape how it answers.
Learn what a model can and cannot do. Tell the context window apart from memory that the application manages, see what reasoning and multimodal models add, and recognise the common errors, from invented facts to answers that change with the wording.
Compare the ways a model reaches your product, a hosted API or open weights on your own hardware. Check licences and data terms, and weigh latency, throughput, memory and quantization at the level of a decision.
Evaluate models on your own task. Build a small representative test set, score answers with a consistent rubric before you see the model names, and measure latency and token use.
Turn the evidence into a decision that lasts. Weigh requirements, plan a fallback for outages and model retirement, and write a recommendation with its assumptions and the triggers to check it again.
Two end-to-end case studies on public and synthetic data. Choose a model for bilingual government answers, and choose a model that must run on your own hardware. Complete at least one for the certificate; each one is a portfolio piece.
Complete at least one case study for the certificate. The others are optional. What each one covers
Use every module in one practice project. You collect evidence, and check it against the course outcomes.
Realistic projects from start to finish, on real data, with every output explained. Each one makes a portfolio piece. Complete at least one for the certificate; the others are optional.
Choose a model for bilingual government answers3 hr
Compare hosted and local models on real English and French Government of Canada questions, with a rubric, measured cost and latency, and a decision memo.
Choose a model that runs on your own hardware3 hr
A privacy rule forbids sending the text to a provider: compare open-weight models and quantizations on a laptop, measure memory, speed and quality, and recommend one.
Other people teach this subject well too. We chose these to go with this course. They open on other sites.See all community picks
Loading the reviews…
Took this course? Sign in to review it. Learners can write a review after they complete 25% of the lessons.
