An AI certificate earned by clicking through a quiz tells a hiring manager nothing. Ours are practical briefs: you build the thing, a working engineer reviews it, and you talk them through your decisions. If you can't explain it, we don't issue it.
Sit them in any order. Most people take the one matching the track they've just finished, while the material is still warm.
Demonstrates you can direct an agent through real engineering work in a codebase you didn't write, and review its output to a standard you'd defend in a pull request.
The senior credential. Design, build, evaluate and operate a multi-tool agent — including the failure handling and cost controls that separate a demo from a system.
The foundation credential: a working retrieval-backed application with typed outputs, sensible cost controls and a test suite that proves it behaves.
For engineers who own their inference stack: deploy, quantise, fine-tune and cost a self-hosted model, and prove the quality trade-offs you made.
Every assessment is a practical brief with a working artefact at the end. There is no multiple-choice paper.
A working engineer reviews your submission and talks it through with you live. You get written feedback either way.
Each credential carries a verification link and the model generation it was assessed against, so it means something specific.
One free re-sit is included. The point is competence, not gatekeeping — most people who re-sit pass.
Models move quickly, so every credential records the generation it was assessed against and stays valid for two years. Renewal is a short delta assessment on what has changed, not the whole thing again.
Individual sittings run monthly. For teams we'll run a cohort and assess everyone together, which is cheaper and considerably more useful.