¶ case study — 01 · solo tags: matching · semantic · privacy
the old engine matched keywords. the new one matches by meaning.
matchpro runs on one opinion: matching is not search. for years the matching leaned on exact rules — keyword overlap and hard filters. that works while everyone shares a vocabulary, and gets brittle the moment they don't. so the matching moved from words to meaning.
"matching is not search. search rewards precision. matching rewards being interestingly close."
from rules to meaning
- match by meaning — compare a profile and a role by what they mean, so close-but-differently-worded fits surface instead of getting missed.
- keep the hard rules — meaning finds candidates; explicit constraints still decide who actually qualifies. the boring rules earn their keep.
- don't lean on one vocabulary — meaning travels where keyword tables don't, so the same approach reaches past a single field or language.
- privacy by default — matching runs over anonymized profiles; identity is revealed only by mutual consent.
where it landed
it surfaces interestingly-close people that keyword matching walked past — while keeping both sides anonymous until they agree to talk. no brittle per-field synonym table to maintain.
still running. still learning.
● matchpro.io · matching · semantic · privacy