The AI Engineering Academy
Where AI actually helps, and where it will embarrass you.
Task-by-task guides for device engineers. Each one names what AI does well, shows a worked example on synthetic data, and is explicit about what it gets wrong — because the failures are the part nobody publishes.
How these are written
Every guide follows the same shape, because the shape is the argument:
- The task, stated as an engineer would state it.
- Where AI helps, and where it fails — with the failure named specifically, not hedged.
- Prompt patterns that work, and why.
- A worked example on synthetic data. Never a real device, never a client's file.
- A verification checklist, because the output is a draft.
- What never to delegate — the judgements that stay yours.
Nothing in these guides produces a DHF-ready document. AI drafts; a named engineer decides and signs. A guide that blurred that line would be teaching the habit that gets a file cited.
Guides
-
Accelerated aging protocols In draft
Where AI genuinely helps with an ASTM F1980-shaped protocol — assembling the rationale structure, cross-checking the arithmetic, catching the assumption you left implicit — and where it produces a confident Q10 justification that will not survive a reviewer.
-
Standards research and applicability mapping In draft
Using AI to build a candidate standards list and an applicability rationale for a device, and the specific failure that matters: it will cite a standard that does not apply, or a clause that does not exist, in a tone indistinguishable from the correct ones.
More are being written. When a guide publishes, the weekly brief carries it first — subscribe here.