Resume Tailor

Step 1: Strategy Planning

Before writing anything, plan the resume strategy:

Keyword Extraction

Pull from the JD:

• Primary keywords: Skills/tools mentioned in must-haves (use these 2-3 times)
• Secondary keywords: Nice-to-haves, methodology terms, industry jargon (use once)
• Action verb alignment: What verbs does the JD use? (built, led, designed, scaled, optimized, partnered) — mirror the energy

Experience Prioritization

Decide which roles and achievements from the profile are most relevant:

1. Rank each work history entry by relevance to the JD (high/medium/low)
2. Rank achievements within each entry by JD alignment
3. Decide how many bullets per role: most relevant gets 4-5, others get 2-3, irrelevant roles get 1 or get cut entirely
4. Determine if any projects should be elevated to their own section

Achievement Triage (When Over Page Limit)

When the resume exceeds the target length, use this scoring framework to decide what to cut. Never cut arbitrarily — every cut should be justified.

Score each bullet on three axes (1-5 each):

RELEVANCE — How directly does this map to a JD requirement?

• 5: Directly addresses a must-have skill or responsibility
• 4: Addresses a nice-to-have or closely adjacent skill
• 3: Demonstrates transferable capability
• 2: Shows general competence but weak JD connection
• 1: No meaningful connection to this specific role

IMPRESSIVENESS — How strong is the impact?

• 5: Quantified, large-scale, clear business outcome ("saved $2M", "served 10M users")
• 4: Quantified, meaningful scope ("reduced latency by 40%", "led team of 8")
• 3: Some metrics or clear scope but moderate impact
• 2: Descriptive but lacks quantification ("improved performance")
• 1: Vague or task-based ("responsible for X")

UNIQUENESS — Does this differentiate you from other candidates?

• 5: Unusual achievement most candidates for this role cannot claim
• 4: Strong achievement that some candidates might match
• 3: Solid but common for this career level
• 2: Expected for someone at this level
• 1: Table stakes — everyone applying will have this

Total Score: 3-15. Cut lowest-scoring bullets first.

Tiebreakers (in order):

1. Keep the more recent role's bullet (recency bias is real in hiring)
2. Keep the bullet that covers a JD keyword not covered elsewhere
3. Keep the bullet with the stronger STAR story backing (for interview prep)

Cut Strategies by Score Range:

• 12-15: Keep — these are your strongest bullets
• 9-11: Keep if space allows, trim wording if not
• 6-8: Cut first, unless it covers a unique keyword
• 3-5: Cut immediately — these are diluting your resume

After triage, verify that keyword coverage hasn't dropped below 70%. If cutting a bullet removes the only mention of a must-have skill, move that skill mention to the Skills section instead.

Section Ordering

Choose the optimal resume structure for this specific application:

• Experience-first (default for 3+ years experience): Summary → Experience → Skills → Education
• Skills-first (for career changers): Summary → Skills → Relevant Projects → Experience → Education
• Education-first (for new grads or PhD-required roles): Summary → Education → Research/Projects → Experience → Skills

Step 2: Writing Each Section

Professional Summary (3-4 lines)

Write a targeted summary that addresses the JD's core needs in the first sentence.

Formula: [Years] + [domain] + [role type] + [with expertise in JD's top 3 requirements]

• [unique differentiator from profile] + [impact statement with numbers]

Example: "Senior data engineer with 6 years building production ML pipelines and
real-time data infrastructure. Expertise in Spark, Airflow, and dbt with a track
record of reducing pipeline latency by 40% and processing costs by $2M annually.
Passionate about data quality and developer experience."

Experience Bullets

Each bullet follows the Impact-First Formula:

[Strong action verb] + [what you did] + [for whom/what scale] + [measurable result]

Rules:

• Start every bullet with a strong, varied action verb (never repeat verbs)
• Include at least one metric per bullet when possible
• Front-load the most JD-relevant bullets in each role
• Naturally incorporate primary keywords without forcing them
• Keep bullets to 1-2 lines each — concise, scannable

Good bullet examples:

• "Architected real-time event pipeline processing 2M+ events/day using Kafka and Flink, reducing data freshness SLA from 4 hours to under 15 minutes"
• "Led cross-functional team of 8 to migrate legacy ETL to dbt, cutting transformation runtime by 60% and eliminating 200+ hours/month of manual data fixes"

Bad bullet examples (avoid):

• "Responsible for data pipelines" (no impact, no specifics)
• "Utilized cutting-edge technologies to drive synergistic outcomes" (buzzword soup)
• "Worked with team on various projects" (vague, passive)

Skills Section

Organize skills to front-load JD matches:

• Group by category: Languages | Frameworks | Cloud/Infra | Data | Tools | etc.
• List JD-matched skills first in each group
• Include proficient+ skills only — don't list things at "familiar" level unless the JD specifically mentions them as nice-to-have

Education

• Include GPA only if >3.5 and graduating within last 5 years
• List relevant coursework only if it directly maps to JD requirements
• Certifications that match JD requirements go here or in a separate section

Step 3: ATS Optimization

Run these checks on the finished resume:

1. Keyword density: Every must-have skill from the JD appears at least once in the resume (ideally in both Skills and Experience sections)
2. No graphics/tables/columns: ATS parsers choke on complex formatting. Use a clean, single-column layout for the primary version
3. Standard section headers: Use "Experience" not "Where I've Made Impact." ATS systems look for conventional headers.
4. File format: PDF preserves formatting; DOCX is sometimes requested. Generate both.
5. No headers/footers for critical info: Some ATS skip header/footer regions — name and contact info go in the body
6. Date formatting consistency: Use "Month Year – Month Year" consistently
7. Spell out acronyms once: "Natural Language Processing (NLP)" — catches both forms

Step 4: Generate the Document

Read the appropriate SKILL.md for the output format:

• For PDF output → read /mnt/skills/public/pdf/SKILL.md
• For DOCX output → read /mnt/skills/public/docx/SKILL.md

Resume Template Structure

[Full Name]
[City, State] | [Phone] | [Email] | [LinkedIn URL] | [GitHub/Portfolio URL]

PROFESSIONAL SUMMARY
[3-4 line targeted summary]

EXPERIENCE
[Most Recent/Relevant Role]
[Company Name] | [Location] | [Start – End]
• [Impact bullet 1 — most JD-relevant]
• [Impact bullet 2]
• [Impact bullet 3]
• [Impact bullet 4 if highly relevant role]

[Previous Role]
...

SKILLS
[Category]: [Skill1, Skill2, Skill3, ...]
...

EDUCATION
[Degree, Field] | [University] | [Graduation Date]
[Relevant coursework / honors if applicable]

CERTIFICATIONS (if applicable)
[Cert Name] | [Issuer] | [Date]

Step 5: Validation & Coaching

After generating, run quality checks:

1. Truthfulness audit: Cross-reference every claim against the profile. Flag anything that stretches beyond what the user actually reported.
2. Keyword coverage analysis (target: 70%+): List every must-have skill/term from the JD. Show coverage:

   KEYWORD COVERAGE: 78% (14/18 must-have terms found)
   ✓ Python — Experience section bullet 2, Skills section
   ✓ Spark — Experience section bullets 1, 3
   ✗ Snowflake — NOT FOUND. Suggest adding to Skills as "Snowflake (familiar)"
     if profile supports it, or address in cover letter.
   ✓ dbt — Experience section bullet 4
   ...
   

   If coverage is below 70%, flag it and suggest where to naturally add missing terms.
3. Length check: 1 page for <10 years experience, 2 pages for 10+. Never more than 2.
4. Readability: No dense paragraphs, consistent formatting, scannable.
5. Gap strategy reminder: For each gap identified in the JD analysis, show how the resume addresses it (or flag if it doesn't).

LinkedIn Sync Recommendations

Generate a separate applications/{role-slug}/linkedin-updates.md file listing:

• Headline adjustments to align with the target role
• Summary tweaks to emphasize relevant experience
• Skills to add/reorder on LinkedIn to match JD keywords
• Any inconsistencies between the tailored resume and current LinkedIn (dates, titles, descriptions) that an interviewer might notice

Present the resume to the user with:

• "Here's your tailored resume. Keyword coverage: [X]%. Here's what I prioritized and why..."
• Any trade-offs made
• LinkedIn sync recommendations if profile data suggests mismatches

Step 6: Save Version

Save the resume metadata to the application tracker in the career profile:

{
  "resume_version": "resume_[company]_[role]_[date]",
  "target_job_id": "jd_xxx",
  "keywords_optimized_for": [],
  "sections_modified": [],
  "achievements_highlighted": []
}

Iterative Refinement

If the user asks for changes:

• "Make it more technical" → increase technical detail in bullets, add architecture specifics
• "Too long" → cut least-relevant bullets and roles
• "Add more leadership" → elevate management/mentorship achievements
• "Make it more senior" → use strategic language, emphasize scope and impact over tasks
• "ATS isn't picking it up" → check keyword coverage, simplify formatting, spell out acronyms

Over-Qualification Strategy

When the user's profile seniority exceeds the JD seniority by 2+ levels (e.g., Director applying for Senior IC, Staff Engineer applying for Mid-level), the resume needs deliberate calibration to avoid the "overqualified" rejection.

Why Overqualified Candidates Get Rejected

Hiring managers worry about:

1. Flight risk: "They'll leave as soon as something at their level opens up"
2. Salary mismatch: "They'll want more than our budget allows"
3. Management fit: "They'll be frustrated reporting to someone junior to them"
4. Culture friction: "They'll try to change how we do things"

Resume Adjustments

Summary: Lead with "why this role" — genuine enthusiasm for IC work, domain passion, specific interest in this company's problem space. Make the downlevel feel intentional, not desperate.

Example: "Staff-level engineer seeking a focused IC role building [specific tech from JD]. After [N] years scaling teams, I'm returning to what I love most: hands-on [domain] work. Looking for depth over breadth."

Title Framing: Consider using functional descriptions over management titles when the management experience isn't relevant:

• Instead of "VP of Engineering" → "Technical Lead, Platform Engineering"
• Instead of "Director of Data Science" → "Senior Data Scientist / Tech Lead"
• Only adjust if the user agrees — never misrepresent. Explain the trade-off.

Achievement Reframing:

• "Managed team of 12 engineers" → "Partnered with 12 engineers to ship X"
• "Set quarterly OKRs for the org" → "Defined technical strategy for [project]"
• "Grew team from 3 to 15" → "Scaled [system] to handle [metric]"
• Keep management achievements only if they demonstrate relevant IC skills

What to Emphasize:

• Recent hands-on technical work (even if it's a small portion of the role)
• Individual contributions within team projects
• Technical decision-making, architecture, code reviews
• Mentorship framed as "technical leadership" rather than "people management"

What to De-Emphasize (not remove — reframe):

• Team size and direct reports
• Budget ownership
• Hiring and performance reviews
• Strategy and roadmap ownership (unless it's a Staff+ IC role)

WARNING: If the user is downleveling by more than 2 levels, flag this for discussion before proceeding: "You're applying for a [level] role with [level] experience. This can work if it's intentional, but hiring managers will ask why. Let's make sure your narrative is airtight. What's driving this move?" The answer becomes the foundation of the resume's positioning.

Tone

Be confident in the resume but transparent about trade-offs. "I emphasized your ML pipeline work because that's 3 of their 5 must-haves. Your frontend experience is real but less relevant here, so I kept it to one bullet." The user should understand every decision so they can override if needed.