Everything an AI engineer meets from linear algebra to multi-agent systems in production, plus the two phases the courses leave out: deploying inside a customer's business, and governing what you shipped. Click any concept to open its lesson. Tick what you know; filter to the forward-deployed essentials when short of time.
The control belongs to the stage. Decide the stage before the control.
Four steps make you an AI engineer. The fifth makes you the one the customer trusts. Use it on every concept below that you have not shipped yet.
Motto, problem, concept. No code. Restate it in two sentences.
The smallest version, framework-free. Type it, do not paste it.
The same with a real library. Match results to your version.
Keep a reusable output: prompt, eval, tool, runbook.
Stage, risk, control, owner, evidence. One row in the register.
Three goals, three stages each. Finish the level 3 project and you have the proof a hiring manager, a customer or a board asks for.
Each manual is a document you will hand to a customer, an auditor or your own team. The fields are the whole manual; fill them and you have the artefact. These are the files that turn "we have governance" into evidence.
Forward-deployed and governance scenarios, one screen each. Pick the best next step, then read why. The AIGP question bank with mock exams lives in the separate practice drill, linked below.
The AIGP practice drill has the core question set, timed mock exams in the real format, a tough mode and a recap tab.
Rate yourself on each skill: 0 never done it, 1 can follow, 2 can do with help, 3 can do alone, 4 can teach it and have shipped it. Be hard on yourself; the gaps are the plan.
Star any concept in the catalogue and it appears here. Use it for the handful you want to revisit this month.
Your own editions first, then the primary sources worth reading in full. Each primary source counts as self-study CPE: one credit per forty pages, up to five per term.
The stack an engineering leader with a governance remit needs: one governance credential, one deployment credential, one standard that connects them. After that, every hour goes into deployments and writing about them.
| Date | Activity | Credits |
|---|
A weekly newsletter on enterprise AI governance, written from inside an engineering organisation that ships AI to regulated customers. Each edition takes one thing that happened, one law that landed or one paper that matters, and turns it into the questions an engineering leader should ask on Monday.
Two running series: Papers from the Trenches, one research paper per edition read for what it changes in practice, and the incident readings, where a public AI failure is traced to the control that was missing.
Published on LinkedIn. Paste the newsletter link below once and this page remembers it in your browser.
What AIGP gets right about enterprise AI governance, and what it leaves to you. The practical version of the certification for engineering teams.
The October disclosures. Agents that paid from wallets they were not given and disabled their own monitors, read as a deployer's checklist.
No. 3: Cemri et al., why multi-agent systems fail (MAST). No. 4: Lynch et al.