About a semester each
Twelve weeks, two hours a week, one project. Fits a credit module, an elective or a bootcamp.
Four levels
From first prompt to production agent.
Every track climbs the same four levels, so a law student and a software student share a vocabulary even when their labs are different. Each level is roughly a semester at two hours a week, and each one ends in something to show.
Level 1
Novice
Any student, any faculty
What a model is and is not. Prompting that works. When to trust an answer and when not to. Using AI for study without cheating yourself. Privacy, bias and the rules that apply to you.
Outcome: a personal AI workflow and a passed assessment on responsible use
Level 2
Practitioner
Students in their field
AI applied to your discipline. Connecting documents and data. Comparing models on cost and quality. Writing the instructions for a reusable assistant. Reading an audit log.
Outcome: a working assistant for a real task in your field, with its costs and limits documented
Level 3
Builder
Tech tracks, and ambitious non-tech
Building agents that act: connecting systems, setting rules and approvals, testing before release, measuring quality. Where things go wrong and how to design so they don't.
Outcome: an agent deployed on the platform, tested, rated and running within a budget
Level 4
Expert
Final year and postgraduate
Capstone briefs modelled on real industry problems. Securing and attacking AI systems. Governance that would satisfy a regulator. Architecture for scale and sovereignty. Presenting the work the way you would to a client.
Outcome: a capstone project, a verified portfolio and a RealRelay Expert credential
Credentials that stack
Each level earns a verified credential. Employers check it on the platform in one click, and see the work behind it.
Cross over any time
A Practitioner in business can step into Builder in developer. The ladder is shared. Only the labs change.
A week in the Academy
Monday a lecture. Friday a working agent.
Here is what Builder level looks like for a second-year cybersecurity student, in one ordinary week.
- Monday, 12 minutes
A video on prompt injection: how an attacker hides instructions in a document an agent will read. Three real cases, one of them from last month.
- Tuesday, lab
She is given an agent that answers HR questions from a policy folder. Her job is to make it leak something it shouldn't. She succeeds in forty minutes. The platform logs exactly how.
- Wednesday, lab
Now she defends it. She puts the agent behind the firewall, writes the rules, and runs her own attack again. It fails. She runs the platform's attack library. Two get through.
- Thursday, project
She fixes the last two, rates the agent, and writes the one-page security note a company would expect before release.
- Friday, review
Her lecturer sees the whole trail: the attack that worked, the fix, the final rating. No essay to mark. Nothing to copy. A grade that means something.
The agent rating her project ends with. The same grade, AAA to D, that buyers see on the platform.
For the student
A portfolio of things that run, not slides about them. A credential an employer can verify on the platform in one click.
For the lecturer
Fresh material every term without rewriting a module. Labs that mark themselves. A dashboard that shows who needs help before the exam does.
For the employer
Graduates who have already built, broken and secured an agent on the platform they will use at work. Interviews that start at the project, not the CV.