AI Engineering
Career changers, engineers adding AI to their work, and graduates who want an AI engineering role rather than a passing familiarity with the tools
A person reads your work
Your capstone projects are not auto-marked. A reviewer reads each one within 10 working days and either approves it or sends back written notes on what to change, and you resubmit. The certificate is signed by hand only after every project has passed.
2 hours reading + 17 hours of projects
AI Engineering is an online course for people moving into AI work. It covers what the field actually is, the technical floor you need, and four capstone projects you keep as a portfolio. Finish all four and a reviewer hand-signs your certificate.
Build four portfolio projects: a web app, a mobile app, a verified data analysis, and an agent with an evaluation report. Learn model behaviour, retrieval, fine-tuning, evaluation, operating costs and job-search preparation. This is an applied course for learners who can already write and debug basic software; it does not replace a programming foundation or guarantee employment. The certificate follows completion of the lessons and approval of all four projects.
What you'll be able to do
- Explain what happens between a user pressing send and an answer appearing, in the terms an interviewer expects
- Choose correctly between prompting, retrieval and fine-tuning, and defend the choice on cost and maintenance
- Build a retrieval system over real documents and detect the retrieval misses that make it lie
- Evaluate an AI feature with a test set instead of an impression, and report where it fails
- Ship four portfolio artifacts: a web app, a mobile app, a verified data analysis, and an agent with an honest evaluation
- Run an AI job search end to end, from reading a job description accurately to the technical screen
Modules in this course
The job families that get called 'AI jobs', including product, platform and architect roles, what each one actually does, and which of them you are realistically a candidate for.
12 min read
The distinction between training and inference, what attention does at intuition level, and what model size does and does not buy you.
9 min read
How text becomes numbers that can be compared for meaning, what that makes possible, and the failure modes nobody warns you about.
9 min read
Ingestion, chunking, retrieval, reranking and injection, how the vector index is tuned, and the silent failure that makes a grounded system lie with total confidence.
13 min read
What fine-tuning genuinely changes, why it teaches form far better than facts, and the decision rule that saves organisations six-figure mistakes.
10 min read
Treating a production prompt as code, not as a lucky sentence: structure, versioning, regression testing, and defending against injection.
10 min read
What a tool call actually is, why multi-step autonomy compounds error, and the constraints that make an agent safe enough to ship.
9 min read
Building a test set for a system with no single right answer, using a model as a judge without fooling yourself, and what to report.
9 min read
How token pricing actually behaves, where latency comes from, and a selection process that survives the next model release.
10 min read
Where user data actually goes, the failure modes that only appear at scale, and the operational habits that separate a demo from a product.
10 min read
How to make a portfolio easy to assess, show measured results and present the four projects in this track.
8 min read
Reading a job advert accurately, the routes that actually produce interviews, and how to approach people without wasting their time or yours.
8 min read
The four stages an AI role typically uses, the questions that separate candidates, and how to talk about work you did with AI assistance.
9 min read
Evaluating an AI role before accepting it, negotiating without theatre, and establishing credibility once you are inside.
8 min read
A deployed web app that answers questions over a document set you supply, cites its evidence, and refuses when the answer is not there.
240 min to build
A mobile app, installable on a real device, whose AI feature is designed around the constraints that make mobile different.
240 min to build
An analysis of a genuinely dirty dataset, where a model writes the analysis code and you build the checks that prove the resulting numbers are right.
240 min to build
A tool-using agent plus a written evaluation over at least twenty cases, documenting where it fails and what you changed in response.
300 min to build
Questions before you start
- How long does this course take?
- About 2 hours of reading, plus roughly 17 hours across the 4 build projects. Most of that time is the build projects, at your own pace.
- Do I need any experience first?
- No formal prerequisites — every course is written to be readable from the first lesson, and that lesson is open to read before you sign up, so you can check it's pitched right for you.
- How is the certificate issued?
- A reviewer reads all of your build projects within 10 working days each and hand-signs your certificate once they're approved — not an automated stamp.
- What does it cost?
- Nothing. The whole course, including the certificate, is free with a Course Parse account. There is no subscription and no card to enter.