The HireScore Algorithm
Skills-Weighted Candidate Compatibility Scoring for the UK IT Market
Abstract
HireScore is the proprietary candidate-job compatibility scoring algorithm at the core of Hirebyai.io. This paper describes how the algorithm works, why it is designed the way it is, and where it is going.
The current version, HireScore V1, operates as a two-stage system. In the first stage, a deterministic matching engine compares candidate skills against job requirements using normalised fuzzy string matching. In the second stage, a large language model evaluates the full candidate profile using structured JSON output enforcement and scores compatibility across three weighted pillars.
This design was chosen deliberately. Skills are weighted most heavily because they are the most objective and verifiable dimension of fit. Experience adds seniority context that skills alone cannot capture. Profile Fit provides a composite signal covering factors that matter in the UK IT market but are harder to score mechanically.
Two-Stage Scoring Architecture
Runs immediately when a candidate applies. Fast, deterministic, and explainable. Every run on the same inputs produces the same output.
Uses normalised fuzzy string matching with a configurable distance threshold to handle real-world variations: React vs ReactJS, Kubernetes vs K8s, PostgreSQL vs Postgres.
Runs asynchronously after application submission. Designed for employers to rank candidates within an applicant pool so the most relevant CVs surface first.
The full anonymised candidate profile is evaluated by a large language model using structured JSON output enforcement across three weighted pillars.
Stage 1 Sub-Weights
Stage 2: The Three Pillars
Direct comparison of candidate skills against required and preferred job skills using normalised fuzzy string matching with a configurable distance threshold.
Evaluates years of relevant experience, seniority progression, and alignment with the role's expected level from junior through to lead.
Holistic assessment of career trajectory, role progression patterns, and general alignment between the candidate's professional narrative and the role context.
Privacy by Architecture
All personally identifiable information is stripped before any AI call. Models only ever receive anonymised structured data.
Any retained PII is encrypted at rest using AES-256-GCM with unique per-record initialisation vectors.
GDPR-compliant automatic cleanup removes expired CV analyses. Data does not persist indefinitely.
CV processing enforces security at the architecture, application, and data layers with no optional bypass.
V2 Roadmap
HireScore V2 expands the Profile Fit pillar into four discrete sub-factors with their own weights, and introduces a closed-loop validation system that tunes weights against actual hiring outcomes.
Postcode-level distance scoring for on-site and hybrid roles.
Range overlap check between candidate expectations and advertised band.
Structured matching for SC/DV clearance, AWS, Azure, CISSP, and UK-specific qualifications.
Formalised progression analysis: title advancement, tenure patterns, and seniority velocity.
Ready to See HireScore in Action?
Book a demo to see how HireScore can transform your recruitment pipeline, or download the full technical whitepaper for the complete specification.