Refine resume tailoring metrics and matching accuracy

[dsmithnautel]

Summary

Improve the core tailoring algorithm to better match resume content to job descriptions and develop metrics to measure effectiveness.

Current Approach

• LLM scores atomic units 0-10 based on JD relevance
• Greedy optimizer selects highest-scored units within constraints
• Coverage tracked as percentage of must-haves matched

Areas for Improvement

Scoring Accuracy

• Improve scoring prompt to better identify relevance
• Weight must-haves higher than nice-to-haves
• Consider recency of experience in scoring
• Factor in keyword density and specificity
• Handle skills synonyms (e.g., JS vs JavaScript)

Matching Logic

• Better requirement extraction from JDs
• Skill taxonomy for smarter matching
• Context-aware matching (senior role needs senior experience)
• Industry/domain relevance weighting

Tailoring Metrics

• Define what makes a well-tailored resume
• ATS compatibility score
• Keyword coverage percentage
• Section balance scoring
• Relevance density (relevant content per page)

User Feedback Loop

• Allow users to rate tailored output quality
• Track which selected bullets users keep vs remove
• A/B test different scoring approaches
• Collect data on interview callback rates (long-term)

Constraints Refinement

• Optimal number of bullets per role
• Section prioritization logic
• Page density optimization
• Skills section curation

Priority

Medium - core to product value proposition

[dsmithnautel]

Additional Scope: Prompt Refinement

Merging scoring/prompt refinement request into this issue:

Prompt Engineering

• Research what makes a strong resume for different roles
• Define clear criteria for good JD matching
• Test different scoring prompts and compare results
• Document prompt templates and reasoning

Good Resume Criteria

• Relevant experience to role level
• Keyword density without stuffing
• Quantified achievements
• Clear action verbs
• Appropriate length and formatting

JD Matching Quality

• Must-have requirements coverage
• Nice-to-have bonus points
• Industry/domain relevance
• Experience level alignment