How HUBCV's skill engine weighs the evidence

Most "resume builder" tools are templating engines: you type in your experience, and it formats it nicely. HUBCV's engine works backwards from that — it starts from a record of things that verifiably happened, scores that evidence, and only then generates the document.

Every piece of activity lands in one of five evidence tiers, and the tiers aren't evenly weighted — deliberately. Completing a course or a job counts far more than a single graded assessment; an assessment or task (scaled by the actual score, so a failed test contributes almost nothing) counts more than attendance; attendance counts more than simply being logged in a room. That hierarchy grew out of a resume-generation method described in a patent application the platform was originally inspired by (US 2020/0074567), and has since evolved into something broader.

The result is a skill score that behaves the way a teacher's or hiring manager's intuition already does: finishing something real moves the needle most, doing the work along the way still counts, and showing up counts a little — but nothing is invented, because every tier is computed from evidence rows that exist in the record.

Once every skill has a score, the resume is generated by selecting and ordering the highest-signal items for the role you're targeting — which is why the same profile produces a different-looking resume for a research assistantship versus a software engineering internship.