The product work behind each project
For each product I publish five artifacts: a problem brief, an evaluation spec, a metrics tree, a decision log and launch notes. They show how each product was framed, measured and decided. Every line links back to the project's own documents.
Which artifacts are published for each project?
5 of 5 projects have published artifacts, with 24 dated decisions logged so far.
What does each artifact show?
Accredible's 2025 research found employers value applied skills in a real-world context (58%) and test or assessment results (51%). Each artifact answers a question a hiring manager or client would ask.
| Artifact | Question it answers | What to look for |
|---|---|---|
| Problem brief | Did they understand the user and the constraints? | A specific user, a job to be done, and non-goals that rule things out |
| Evaluation spec | How is "good" defined for a probabilistic product? | A test set grouped by failure type, metric floors, and when floors must be re-derived |
| Metrics tree | What is the product trying to move, and what must not break? | One north star, the drivers under it, and guardrails |
| Decision log | How do they decide under uncertainty? | Dated decisions with evidence, including rejected options and overrides |
| Launch notes | Are they honest about what shipped? | Shipped scope, known limits and the next step |
What do people ask about product artifacts?
What are product artifacts in an AI PM portfolio?
The documents a product manager produces while framing and shipping a product: a problem brief, an evaluation spec that defines quality, a metrics tree, a decision log and launch notes. For AI products the evaluation spec matters most, because it is how "good" is defined for a probabilistic system.
Why show artifacts instead of only results?
Employers look for applied skills in a real-world context and for test results. Artifacts show the judgement behind the results: what was measured, what was rejected and why, and what is still weak.