Context
Roster management requires reviewers to compare resumes against labor categories, experience requirements, credentials, and role-specific skills. The process can become inconsistent when reviewers must interpret many resumes across multiple job classes and programs.
Challenge
The project needed an AI workflow that could support screening and role matching while preserving explainability and reviewer control. It also needed to handle personally identifiable information responsibly and avoid turning staffing decisions into an opaque black-box recommendation.
My Role
I led the design and development of the resume-to-job matching workflow. I created the scoring structure, prompt logic, PII handling approach, and local LLM analysis process.
Approach
The workflow uses AWS Comprehend for PII removal and locally hosted LLMs through Ollama for content analysis. It evaluates resumes against multiple labor categories using weighted scores for education, field of study, certifications, skills, total experience, and relevant experience. The system also produces match reasoning and identifies missing or concerning items.
Output
The project produced an internal AI tool that recommends best-fit roles, generates relevance scores, explains match logic, and supports comparison across labor categories.
Impact
The tool improved consistency and transparency in staffing review. It helped reviewers focus on judgment and validation rather than repetitive extraction and comparison work.