abhishekverma73072 / resume-tailor

Tailoring a student resume to a job description

Goal: make the truth easier for a recruiter to see. Never make the resume say
something the student has not done. A resume that gets shortlisted and then
collapses in the interview is worse than one that is not shortlisted.

When to use this

Use when the user provides (or can paste) BOTH a resume and a target JD. If one
is missing, ask for it before doing anything else. Do not tailor to a vague
role name like "software job"; ask for the actual JD or at least the company
and role.

Hard rules (never break these)

1. No invented facts. Do not add skills, tools, projects, employers,
   numbers, grades, dates or certificates that the student did not state. If a
   JD asks for something the resume lacks, report it as a gap. Do not hide it
   by wording.

2. No invented metrics. If a bullet has no number, do not make one up. Ask
   the student ("How many users? How much faster? How many records?") and use
   only what they answer. If they do not know, write the bullet without a number.

3. Rephrasing is allowed, upgrading is not. "Helped with a Flask API" must
   not become "Led backend development". Keep the student's real level of
   ownership (built, contributed to, assisted).

4. Keep personal data as given. Do not change name, contact details, CGPA,
   percentages or dates. Do not echo phone numbers or email addresses in your
   analysis; refer to them as "contact line".

5. Do not copy phrases from the JD word for word into bullets just to match
   keywords. Use the JD's terminology only where it truthfully describes
   what the student did.

Workflow

1. Read both documents

Get the resume text (paste, or file; for PDF or DOCX extract the text first).
Read the JD and split its requirements into three groups:

• Must-have: stated as required, "must", "should have", or repeated.
• Nice-to-have: "preferred", "plus", "good to have".

• Eligibility filters common in Indian campus hiring: minimum CGPA or
  percentage, branch, graduation year, backlogs allowed or not, location,
  bond or service agreement. These are pass or fail; see step 3.

2. Run the keyword gap check

If a shell is available, run:

python3 scripts/keyword_gap.py --resume resume.txt --jd jd.txt

It prints which JD terms already appear in the resume and which do not. Treat
the output as a lead, not a verdict: it does not understand synonyms
("REST API" vs "web services") and it counts words, not experience. Read the
resume yourself and correct it. If no shell is available, do the same check by
reading.

3. Check eligibility first

Compare the filters from step 1 with what the resume states. If the student
clearly does not meet one (for example a CGPA cutoff), say so plainly and
early, before spending effort on wording. Do not suggest hiding or rounding
up a grade. Suggest they confirm the cutoff with their placement cell, since
many companies apply it only at the shortlisting stage.

4. Map evidence to requirements

For each must-have, find the strongest true evidence in the resume: a
project, internship, coursework, certification, competition. Produce a table:

Strength is Strong (shipped or used it), Partial (touched it, or only
coursework), or Gap (nothing). Action is "move up", "reword", "add detail
(ask student)", or "gap: be honest, see step 7".

5. Rewrite

Change only these things:

• Order. Put the most relevant project or experience first. Within a
  project, put the most relevant bullet first.

• Wording. Each bullet: strong verb + what was done + tool/tech + result
  if known. One line, ideally two at most. Past tense for finished work.

• Skills section. Reorder so JD-relevant skills come first. Remove nothing
  true unless it is clutter; do not add anything untrue.

• Summary or objective. Optional for freshers. If included, two lines,
  naming the target role and two to three true strengths. Skip it if it would
  be generic.

• Length. One page for students and freshers. Cut the weakest items before
  shrinking the font below 10 pt.

Keep the format simple: single column, standard headings (Education, Skills,
Projects, Experience, Achievements), no tables, icons, photos or text in
images, so that applicant tracking systems can read it.

6. Ask for what is missing (once, together)

After drafting, ask the student for the missing detail in a single short list:
numbers for bullets that have none, links (GitHub, deployed demo), and
whether an item you were unsure about is real. Do not ask one question at a
time.

7. Handle gaps honestly

For every must-have gap, give one of:

• an honest framing of nearby real experience ("Used PostgreSQL in the
  college DBMS project; no production MySQL experience" is fine to state in
  an interview, not necessarily on the resume);

• a realistic step the student can take this week (a small project, a
  free course with a certificate) and a reminder that it only goes on the
  resume once it is actually done.

8. Deliver

Return, in this order:

1. Verdict: two or three lines. Is this role a realistic fit? What is the
   single biggest issue?

2. Eligibility check (only if there is a filter to report).
3. Requirement-to-evidence table from step 4.
4. Tailored resume as clean text the student can paste.

5. Change log: a short list of what you moved or reworded and why, so the
   student can verify nothing was invented.

6. Questions from step 6 and gap advice from step 7.

If the user asks for a file, produce a .docx or PDF with their resume content
only after they confirm the text.

Common mistakes to avoid

• Keyword stuffing: a skills list of forty tools reads as dishonest.

• Rewriting every bullet. Change what matters for this JD; leave strong
  bullets alone.

• Adding "responsible for", "worked on" or "familiar with" as filler.
• Tailoring to a company's famous tech rather than what the JD asks for.
• Telling a student the resume is "great" when a must-have is missing.

Reference

See references/bullet-patterns.md for before-and-after bullet examples and
examples/ for a worked example.