The short version
TrySignalHire matching starts after both sides provide real context. A candidate submits an evidence-backed packet and chooses public directory discovery. A company creates a profile and opens a job with the role signals it wants to review.
The product then compares the company's open job against submitted, listed candidate packets. AI helps organize the overlap into concise, source-backed match signals, gaps, and follow-up questions. App code calculates a score. A human still reviews the packet, sources, and gaps before deciding what should happen next.
That score is not an auto-reject, auto-hire, ranking guarantee, or employment decision. It is a review aid.
What has to exist first
Three things need to be true before a company match appears:
- The candidate profile is submitted.
- The candidate chose public directory discovery, not just an unlisted packet link.
- The company has at least one open job with a real title and job context.
For the company match queue, TrySignalHire uses the selected job when one is requested. Otherwise it falls back to the company's first open job. Candidate-created role checks in /candidate/applications are different: they are private candidate-side checks and do not create company match rows.
What gets compared
The matching input is intentionally limited to public candidate packet context and job context.
Candidate-side inputs can include the headline, target role, desired roles, location preferences, skills, strongest work example, work history, project evidence, proof links, GitHub, portfolio, LinkedIn, work style fit, and company fit.
Company-side inputs can include the job title, team, location, employment type, seniority, summary, responsibilities, requirements, and evidence signals.
The point is not to compare every private thing a candidate has ever written. The point is to compare the evidence a candidate chose to make reviewable against the signals a company said matter for the job.
What AI does
AI is used to structure the comparison. It receives the candidate/job data that the app passes into the matching request and is instructed to:
- Use only the provided candidate and job data.
- Avoid inventing facts.
- Tie each matched signal to a supported source field.
- Use proof URLs only when the candidate provided a public URL.
- Put uncertainty into gaps instead of pretending the evidence is stronger than it is.
- Avoid private data such as compensation, work authorization, sponsorship, references, raw notes, or private resume storage details.
The AI output is a structured extraction: summary, matched signals, gaps, and follow-up questions. The app sanitizes that output before showing it.
What app code scores
The score is calculated by app code after the signal extraction. Today it is intentionally simple:
- Evidence overlap: up to 40 points for job signals that match source-backed candidate evidence.
- Preference overlap: up to 25 points for candidate preferences that overlap with the job.
- Public proof: up to 20 points for public source links available for review.
- Profile completeness: up to 15 points for usable public profile sections.
Those weights will probably change as the product gets real usage. The important part is that the score is explainable. It comes with source links, matched signals, gaps, and reasons, not just a number.
What humans see
The company review surface shows the candidate packet, match score, source links, matched signals, gaps, follow-up questions, and a 24-hour response window.
The response window is there because silence is part of the problem TrySignalHire is trying to reduce. It does not guarantee a response yet. It makes the pending state visible so the workflow can become more accountable over time.
The company can mark a review as interested or passed. That review state is stored separately from the candidate packet.
What stays private
TrySignalHire is not trying to turn every candidate input into public matching fuel. Raw notes, private resume files, private uploads, reference contact details, salary details, work authorization details, and private candidate-created role checks are not intended to become public directory or company queue content.
Candidate packet links are explicit share surfaces. Public directory discovery is a separate choice. Candidate-created role checks stay on the candidate side and do not create public company pages or company match rows.
Current limits
This is still early. Matching can miss context, overweight weak overlap, or underweight a real fit that needs human judgment. AI output can be incomplete or wrong. The scoring model is a first version, not a settled truth.
TrySignalHire does not make final hiring decisions, reject candidates, guarantee interviews, guarantee offers, or replace human review. The job of the matching layer is narrower: make real work easier to inspect before the hiring funnel turns people into thin resume fragments.