cosmik.work · 10 ~ Hiring

An honest fit score, measured not guessed.

This tool embeds the job description and the resume with a real sentence-transformer model and compares the vectors, before any AI is allowed to opine. The fit number is math. What it means for you is judgment, from one Claude call that sees the score, not the other way around.

The full hiring lifecycle

Sourcing
Before you screen, you have to find. A JD becomes search strings and candidate personas.
→
JD Optimizer
A vague JD poisons every downstream stage. Tighten and de-bias it first.
→
Screening live
The honest, embedding-scored core: fit that can't be talked up or down.
→
Interview Kit
A score tells you whether to talk. This tells you what to ask.
→
Comparison
One role, many candidates, ranked on the same honest measure.
→
Onboarding
Hiring doesn't end at yes. A 30/60/90 that starts day one.
→
→ and someday, one Hiring OS.
Every stage above is one seed of that.

Before you paste a resume: names and contact details are sent to the AI unscrubbed (scrubbing would break the analysis), though never stored beyond the generated result. Make sure that's acceptable to whoever's resume this is.

0 / 30000

0 / 30000

Reminder: unscrubbed candidate details are about to be sent to the AI. Processed transiently, never stored beyond the generated result.

First run can take up to 3 minutes while the embedding model loads. It's faster after that.

Working…
0% Fit

Thresholds: 60%+ Strong fit · 40–59% Moderate fit · below 40% Weak fit.

Computed by comparing sentence-embedding vectors of both documents (cosine similarity). Not generated by an AI, measured.

Strengths

    Gaps

      Interview questions