Overview

Resume Matching with Rule-Based Solutions

Resume matching via rule-based solutions offers a deterministic framework for ranking candidates. By applying predefined logic to resume data and job specifications, the system automatically identifies top talent by combining:

Large Language Models (LLMs) for resume understanding and structured data extraction.
A configurable, rule-based matching engine for transparent and controllable decision-making.
Score-based ranking to surface the best matches with clear reasons.

How It Works

The Resume Matching System bridges the gap between unstructured resumes and intelligent, explainable matching decisions. CubeDrive rule-based resume matching process follows these steps:

Resume Parsing

Resumes are converted into structured data by extracting fields such as skills, experience, education, certifications, job titles, and years of experience.

Rule Definition

Hiring teams define matching rules based on job requirements, for example: Required vs. optional skills; Minimum years of experience; Education level or certifications etc..

Rule Evaluation

Each resume is evaluated against the defined rules. Rules can be: Boolean; Weighted; Conditional etc..

Scoring and Ranking

Candidates receive a matching score based on rule fulfillment and weights, enabling automatic ranking or pass/fail filtering.

Matching Rules

Matching Rules: Clear, Explainable, and Configurable

The core strength of the system lies in its rule-based matching model, which is divided into three logical layers.

1: Hard Rules Mandatory Filters

Hard rules determine eligibility. If a candidate fails any hard rule, they are immediately excluded before any scoring takes place.

Key Examples:

Candidate’s highest degree must be equal to or higher than bachelor degree
Candidate must share at least one common language
Candidate must have specific professional domains

2: Score Rules Ranking Logic

Once a candidate passes all hard rules, score rules determine how good the match is. Each rule defines a specific condition and assigns a quantitative value.

Scoring Breakdown:

Specific city → +20 points
Technical skill match → +25 points
Professional background similarity → up to +25 points
Research field alignment → up to +35 points

3: Business Rules Domain-Specific Enhancements

Business rules allow organizations to inject context-specific logic without changing core matching behavior, making the system highly adaptable.

Context-Specific Examples:

Similar work experience (vector similarity ≥ threshold) → +30 points

Adapts to: Different industries, Employers, or Advisors.

Rule Configuration

Rule Configuration via YAML

All matching rules are defined in external YAML configuration files, not hard-coded logic.

matching_rules.yaml


		- name: "shared_alumni"
		  description: "Teacher and student attended the same university"
		  condition: "student.alumni_id == teacher.alumni_id"
		  weight: 20
		  type: "bool"
		
		- name: "research_alignment"
		  description: "Alignment based on research field keywords"
		  condition: "jaccard_similarity(student.research, teacher.research)"
		  weight: 35
		  type: "compare"

Agility

Business teams can adjust weights and behavior without a developer deployment.

Scalability

Add new rule categories easily as the platform grows.

Versatility

Run different matching scenarios (Teachers, Employers, Advisors) using the same engine.

Conclusion

This Resume Matching System combines LLM-powered data extraction with a declarative, rule-driven engine to deliver accurate, explainable, and customizable matching results. It is designed not just to find matches—but to justify them clearly, adapt to evolving business rules, and scale across different use cases.