Architect a developer-portal personalization engine

WHAT THIS TESTS: It evaluates whether you can architect attribute-driven personalization that stays maintainable as the number of roles, levels, and pieces of content grows over time.

A GOOD ANSWER COVERS: Start with a user profile data model storing attributes like primary role, for example front-end, backend, or DevOps, experience level, languages, and goals, captured explicitly during onboarding and refined implicitly from behavior such as which tutorials they complete. Tag content with the same controlled taxonomy: each tutorial carries role, level, and topic tags plus prerequisites, so content and users speak one shared vocabulary. A matching and ranking layer scores each candidate tutorial against the profile, boosting role and level matches, demoting content the user already finished, and respecting prerequisite ordering. The rendering pipeline pulls the ranked list, applies layout, and personalizes sections such as recommended next and for your role, ideally with edge caching keyed by profile segment and a safe default feed when the profile is still sparse.

COMMON WRONG ANSWERS: Building one hardcoded page per role, which explodes combinatorially and cannot combine role with experience level. Personalizing through template if-statements scattered through the codebase. Providing no fallback for users with empty profiles. Tagging content ad hoc with no shared taxonomy, so the matching becomes unreliable and inconsistent.

LIKELY FOLLOW-UPS: How do you bootstrap a brand-new user's profile before they have behavior. How do you keep content tags consistent across many authors. How do you cache personalized pages without serving stale content. How do you measure whether the personalization actually improves engagement.

ONE CONCRETE EXAMPLE: A new user picks DevOps and intermediate during onboarding. The matching layer surfaces tutorials tagged devops plus intermediate, like a CI pipeline guide, while hiding beginner front-end content. As they finish modules the system marks them complete and promotes the next prerequisite-satisfied tutorial, all driven by tags and rules rather than a bespoke DevOps landing page that someone must hand-maintain whenever the catalog changes.