Methodology
Cofeeds scores candidates with a pipeline of inspectable stages: structured evidence, AI-assisted dimension scoring with rationale, a role-weighted composite, and consent-gated access. The output is a review aid for humans, not an automated hiring decision.
The pipeline
Every stage produces an artifact your team can review. Nothing enters a score without a source, and nothing leaves the pipeline as an unexplained number.
CVs are parsed into structured experience, skills, and education. Voice interviews are transcribed in full. Work preferences are captured as structured fields — so every later step can point back to a specific line, moment, or answer.
Source-linked evidenceAn AI evaluation model scores the evidence across role-relevant dimensions — role fit, skills, communication clarity, and motivation — and writes a rationale tied to CV context, transcript excerpts, and preference answers.
Scores with rationaleEach role sets how much every dimension matters. The composite is arithmetic your team can check — dimension score × weight, summed — not a black box that emits a ranking.
Weighted fit scoreRole requirements and candidate profiles are compared as embeddings to surface relevant candidates. Ranked previews stay anonymized, and identifying detail opens only through an explicit unlock.
Consent-gated shortlistWorked example
Dimension weights are configured per role — these are example defaults, not global constants. A candidate scoring 87 on role fit, 88 on skills, 84 on clarity, and 84 on motivation lands at a composite of 86.
Score math
Frontend Engineer
Role fit
87 × 35%
30.45
Skills
88 × 30%
26.40
Clarity
84 × 20%
16.80
Motivation
84 × 15%
12.60
Composite
86.25
Shown to teams as 86.
Capability status
Cofeeds separates live product capabilities from governance controls and roadmap work. This avoids overstating the AI architecture while keeping every public claim reviewable.
Candidate profiles store structured resume, skills, experience, education, preferences, and account context.
Completed interviews can be transcribed, scored, summarized, and attached to the candidate profile for review.
Employer role context and candidate profiles can be matched, with personally identifying details gated behind employer unlocks.
The product posture is to connect scores back to CV, transcript, and role criteria evidence; public copy should stay aligned to what the interface exposes.
Score behavior, edge cases, and employer workflows are reviewed as governance controls rather than described as fully automated bias certification.
English/French parity is checked across public routes, metadata, legal pages, and key hiring workflows.
Specialized automated reviewers for compliance, hallucination checks, and unsupported claim detection should be treated as roadmap unless tied to production logs.
Human-readable explanations are planned to become more exportable for candidate access, correction, and review requests.
Safeguards
The scoring workflow is designed for review, objection, adjustment, and human accountability before an employer acts on a candidate profile.
Scores point back to profile fields, resume sections, interview transcript excerpts, role criteria, and employer settings rather than unsupported model rationale.
Employers can configure role criteria and adjust which requirements matter most, while the platform keeps a consistent methodology for candidate comparison.
Cofeeds reviews score behavior, sensitive attributes, selection patterns, and edge cases so automated hiring support does not become an unexamined decision-maker.
Employers remain accountable for review and hiring decisions. Candidates can request information about automated assessments and ask for human review.
Candidates see profile status, interview completion, and visibility controls. Detailed employer ranking logic may be summarized where needed to preserve integrity and privacy.
Key events such as profile creation, interview completion, employer unlocks, consent, account changes, and deletion requests are designed to be reviewable.
Employer onboarding
The employer onboarding path captures role creation, criteria weights, preview matches, billing expectations, consent disclosure, and next-step confirmation before full profile access.
