What runs between your resume and your ranked results
NearSkill is a candidate intelligence platform for specialized tech and AI work. It parses a resume into an anonymous skills profile, converts that profile into a vector, and ranks every live opening by similarity against a structured role rubric. Matching runs on NearSkill. Final applications complete on the employer's own system.
The pipeline, in order
Five stages run before you see a single result. Each one is disclosed below with the limits it actually enforces.
You upload a resume
PDF, DOC, or DOCX up to 3 MB. No form, no email, no account.
The file is inspected by content, not by extension. We read the actual bytes to confirm the file is a PDF or Word document, so a renamed executable or script is rejected at intake rather than stored.
If the same resume has been uploaded before, the existing profile is reused and results return immediately. We match on the document itself, not on your identity.
We build an anonymous profile
Skills, seniority, domains, years of work, and tooling.
Text is extracted from the document, then normalized into structured fields. A line that reads "Sr. Backend Eng, 2019 to present, payments infra" becomes discrete values the matcher can compare: role family, seniority band, domain, and date range.
Your name, email address, and phone number are parsed but held back from the matching profile. They reach an employer only when you apply to that employer’s role, and never before.
Every role is rewritten into a rubric
One fixed schema, 21 fields, applied to every opening.
Employers describe the same job in completely different words. Before a role is matchable, an enrichment pass pulls the posting apart and rewrites it against the same schema every other role uses.
The rubric covers identity, place, compensation, requirements, and guidance. Missing pay stays visibly missing rather than being filled with an estimate, because a fabricated number is worse than a blank field.
Your profile is embedded and scored
Vector similarity against every live rubric, ranked by fit.
Your profile becomes a vector. Each role rubric has its own vector. Matching compares them by cosine similarity, which measures how closely two sets of attributes point in the same direction.
The score you see is that similarity expressed as a percentage, rounded to a whole number. It reads only from the structured rubric, never from how well a posting is written.
You apply, and the route is disclosed
Three routes, each labeled on the role before you click.
Some roles NearSkill processes. Some it logs while the employer finishes the form. Some the employer runs entirely. The role page states which one applies before you start, so you always know who receives your data.
No route puts you on a recruiter call list. Running a match never triggers an application.
The rubric every role is rewritten into
Twenty-one fields across five groups. The same shape applies whether the employer supplied the detail or not, which is what makes two openings from two companies comparable at all.
- Identity
- TitleCompanyEngagement typeOpeningsDate postedDemand signal
- Place
- LocationRemote, hybrid, or on-siteEligible countries
- Compensation
- Hourly rangeMonthly rangeAnnual rangeCurrency
- Requirements
- Skill tagsCategoriesRequirementsResponsibilities
- Guidance
- Role overviewWho should applySalary insightApplication tip
The fit score reads from these fields and nothing else. Formatting, tone, and length never enter the calculation.
Where your application completes
NearSkill is not the employer and does not decide who gets hired. What differs between roles is how much of the application we handle before it reaches the employer, and every role states its route before you apply.
| Route | What happens | Who holds the data |
|---|---|---|
| NearSkill processes itRoles NearSkill hosts directly | Logged on NearSkill. Screened against the role rubric. Qualified profiles reach the employer. | NearSkill holds your application and shares it with the employer for this role.Logged on NearSkill. Screened against the role rubric. Qualified profiles reach the employer. |
| NearSkill logs it, employer completes itRoles with a partner application system | Logged on NearSkill at the moment you apply. The employer’s own form finishes the submission. | NearSkill keeps the application linked to your profile. The employer runs the final step on their system.Logged on NearSkill at the moment you apply. The employer’s own form finishes the submission. |
| The employer runs itRoles managed entirely on an external portal | NearSkill opens the employer’s application page. Nothing is captured on our side. | NearSkill collects and stores nothing on this route. The employer receives your details directly.NearSkill opens the employer’s application page. Nothing is captured on our side. |
Roles that NearSkill screens are checked against the same rubric the fit score uses. The screen decides whether a profile moves forward, not whether you are hired. Final decisions belong to the employer on every route.
What NearSkill will not do
- We do not estimate pay
- A role with no published range shows an empty compensation field. No model fills it in, and no figure is derived from similar postings.
- We do not decide hires
- Screening narrows a field. Interviews, offers, and rejections are the employer’s call on every route, including roles NearSkill processes.
- We do not sell resumes
- Resume data is not sold, rented, or shared with recruiters. It reaches an employer only through an application you chose to submit.
- We do not apply on your behalf
- Running a match produces a ranked list and nothing else. No application is created until you submit one yourself.
Common questions
How does NearSkill turn a resume into a fit score?
The resume is parsed into structured fields: skills, seniority, domains, years of work, and tooling. That becomes a vector. Each live role has its own vector, built from its structured rubric. The fit score is the cosine similarity between the two, expressed as a percentage, so it reflects attribute overlap rather than keyword count.
What does a fit score of 70% mean?
It means your parsed profile and that role’s rubric overlap heavily on the attributes both contain. Above 70% is a strong match. Above 50% is worth reading. A low score means the role asks for attributes your profile does not show, not that you would be rejected.
How long does matching take?
Under 20 seconds for a first upload, including parsing, profiling, embedding, and scoring against every live role. Re-uploading the same resume returns results faster because the profile already exists and is matched on document content rather than identity.
Where does my application actually go?
It depends on the role, and the role page says which route applies before you click. NearSkill processes some roles end to end, logs others while the employer completes the form, and for a third group simply opens the employer’s portal while collecting nothing.
Does NearSkill share my name and contact details with employers?
Only when you apply to that employer’s role. Matching runs on an anonymous profile that excludes your name, email address, and phone number. Those details stay server-side until you choose to submit an application, and then they go only to the employer you applied to.
Do you estimate salary when a role publishes no pay?
No. Positions without a published range are counted separately and left blank in the compensation fields. We do not model, infer, or backfill a number. A visible empty field is more useful than a plausible figure that turns out to be wrong.
Data handling is documented in the Privacy Policy. How pay figures and platform research are compiled is documented in our Editorial Standards. Reviewed .
See the pipeline run on your own resume
No account, no email, and no application created by running a match. Upload a resume and read the ranked list.