How it Works
How the matching
actually works.
No black boxes. No buzzwords. Here's exactly what happens when you post a job or create a profile on SeaPlatformJob.
First: what our AI does — and what it doesn't.
✓ What it does well
Converts unstructured CV text and job descriptions into comparable skill profiles. Scores candidates on maritime-specific criteria (certifications, vessel type, fuel systems). Factors in working style and culture preferences. Ranks and explains its reasoning.
✗ What it can't do (yet)
It cannot predict performance. It cannot replace a good interview. It doesn't know if someone is reliable or a team player — only you can judge that. It improves with every rated placement, but right now it has limited data. We'll tell you when that changes.
The baseline it beats
Traditional keyword search, which misses qualified candidates with different vocabulary. LinkedIn's general algorithm, which doesn't understand STCW, vessel types, or maritime culture. Gut-feel hiring, which costs the industry billions annually in bad fits.
How it improves
After every placement, both the employer and the candidate rate the match quality. This feedback is anonymous and feeds directly into the model. The more placements we make, the better the ranking becomes. We publish these scores publicly every month.
Step by step
From signup to placement — here's exactly what happens.
Build a structured profile — not just a CV upload
When you sign up, you don't just upload a PDF and disappear. Both candidates and companies complete a structured profile that captures what traditional CVs miss.
The matching engine builds compatibility vectors
Every profile is converted into a mathematical representation — a "vector" — that captures skills, experience, and personality dimensions. This happens automatically using language models trained on maritime data.
Matches are ranked — with explanations
You don't get a list of 200 applicants. You get a ranked shortlist of the top matches, with a plain-language explanation of why each candidate was ranked where they were.
You decide — the AI advises, not replaces
After reviewing matches, you initiate contact through the platform. After a placement is made, both sides rate the match quality (1–5 stars, plus qualitative feedback). This data improves every future match on the platform.
Under the hood
For those who want to know what technology is actually running this.
NLP & Entity Extraction
Job descriptions and CVs are processed using large language models to extract structured data: certifications, vessel types, skills, and experience duration. No manual tagging required.
Vector Similarity Search
Candidate and job profiles are embedded as high-dimensional vectors. Similarity is computed across skill, culture, and experience dimensions. Ranking weights are tuned by placement outcome data.
10-Question Psychometric Snapshot
Not a full personality test — a short, validated questionnaire that captures working style preferences. Results are converted to vectors and included in the match calculation, weighted at ~20%.
What we don't claim
We prefer honesty over a sales pitch.
We don't claim to eliminate bad hires
We reduce the risk of skill mismatch and culture misalignment. Whether a hire works out still depends on management, onboarding, and many factors we can't see.
We don't have years of data yet
The algorithm improves with placements. Right now we have early data. We'll tell you when the dataset grows large enough to make statistically robust claims.
We're not a replacement for interviews
Our job is to get the right shortlist in front of you faster. The interview and final decision are yours — and should be.
We don't know everything about maritime
Certain niche vessel types, regional manning conventions, and flag state specifics are areas where our data is thin. We flag this when it affects a match.
Early placement results
These numbers are from our first cohort of placements. We update them monthly. View full transparency report →
Ready to see it in practice?
Post your first job free, or create a candidate profile in under 10 minutes.