Finding translators

There are two places to find someone to work with. The people directory is the platform-wide search: translators and project managers, discoverable by language, service, domain, and location. The project translator search lives in a project’s assignment panel and is where you actually staff a seat — it knows the project’s language direction and workflow, so it can order candidates by fit for that specific job. Both use the same ordering rules, described below.

The people directory

The directory lists people on the platform and links to their public profiles. Narrow it down with any combination of filters:

• Name — free-text search.
• Languages — add one or more; chips show what is active.
• Country — add one or more countries.
• Service — translation, proofreading, editing, MT post-editing, transcreation, subtitling, or terminology review.
• Domain — subject-matter area (legal, medical, software, and so on).

Each result shows the person’s name, location, and languages. Click through to their public profile to learn more — the languages they work in and how fluently, the language directions they have declared (with years of experience), the services they offer, and their domain experience — before deciding to send an offer.

Results load in pages; the count above the list shows how many people match.

How results are ordered

The ordering is designed to be fair and predictable, and its mechanics are published here deliberately. Two rules frame everything:

• Filters narrow, ranking orders. A filter removes people who don’t match it; the ranking only sorts whoever remains. Ranking never hides anyone.
• Scoring is additive. Every candidate starts from their language fit, and other signals can only add to that — a signal that doesn’t apply contributes nothing and can never push someone down.

The score is built from four signals:

1. Language fit — how fluently the candidate reports the languages involved. Fluency in the target language (what they produce) counts more than fluency in the source language (what they read), and a native target language earns a boost.
2. Direction — experience in the specific source-to-target direction. A declared language pair on the profile — with its years of experience — always scores higher than merely listing both languages separately. Direction only applies in project search, where there is a real direction; the directory has none, so it contributes nothing there.
3. Service fit — how well the candidate’s offered services cover what the project’s workflow needs.
4. Domain fit — declared experience in the project’s subject-matter domain, when the project has one.

In project search, two groups are placed ahead of the scored order: anyone who already has an offer out for the seat, and — on organization projects — members of that organization.

The fairness mechanics

Merit alone would let the same few profiles sit on top forever, so two counterweights are built in:

• Near-equal scores are ties, rotated daily. Candidates whose scores are within a narrow band of each other are treated as equals, and the order within the tie rotates every day. Being marginally ahead does not buy a permanent lock on the top position.
• Two page-one slots are reserved for less-exposed translators. On the first page of results, up to two slots go to candidates who would otherwise rank below it — chosen from those who have had the least recent exposure, so newer translators and those who haven’t surfaced lately get a real chance to be seen. The selection is stable within a day and rotates daily.

Scores are never displayed — not to PMs and not to translators. What you see is only the resulting order.

If you are a translator, the practical takeaway is that a complete profile is what ranks: declare your language pairs (directions) rather than just listing languages, record your years of experience per pair, and fill in your services and domains. Each of these feeds a signal that can only help.

Auto-place

For PMs who prefer it, each project has an opt-in auto-place convenience, off by default. When enabled, it composes an ordinary offer to the top-ranked candidate for each unfilled seat — using exactly the same ordering as the list, fairness slots included. The offers it creates are normal, cancellable offers that the recipient accepts or declines like any other: auto-place never assigns anyone directly, and you stay in control throughout. See Offers & assignment for what happens once an offer is out.