In my pdfLaTeX sample, pypdf returned two bullets with all their words stuck together. A resume you keep in LaTeX, Typst or Markdown can look fine in the exported PDF, and an applicant tracking system that extracts the PDF's text can still get those words without spaces. I rendered the same sample resume with a Quarto format that typesets the resume with Typst and with the atsresume LaTeX class on CTAN, then pulled the text back out with pypdf and pdfminer.six to see where that happens and what else changes.
What the layout avoids
In both one-column templates, the city sits across from the employer or school, and the dates sit across from the title or degree. The name is centered, with a thin rule under each uppercase heading. There are no tables, text boxes or icons.
A two-cell table can make a PDF extractor separate the city from the company name. atsresume uses stretchable space (TeX glue) to push the city or date to the right margin on the same line. It disables the f-ligatures (fi, fl) and hyphenation, uses the Unicode bullet • for lists, and keeps each heading with the text that follows. It works with pdfLaTeX, LuaLaTeX and XeLaTeX on US Letter or A4, needs only packages from TeX Live's small scheme, and sets the text in TeX Gyre Pagella.
An entry is one command with four arguments:
\experience{Harlan Quay Capital Partners}{New York, NY}
{Investment Banking Analyst Intern}{Jun 2025 -- Aug 2025}
The Quarto version
If you'd rather write Markdown, the Quarto format gives you the same design. It needs Quarto 1.7 or later, and Typst comes bundled with Quarto, so there's no LaTeX install:
quarto use template supertj/quarto-readable-resume
Each entry is a div whose details are attributes, with the bullets inside:
::: {.experience
company="Harlan Quay Capital Partners"
location="New York, NY"
title="Investment Banking Analyst Intern"
date="Jun 2025 – Aug 2025"}
- Built three-statement models for two consumer retail clients
:::
There are three entry types (.education, .experience, .activity), and every attribute is optional. The detail I like: a misspelled attribute stops the render with an "unknown argument" error, so a typo in locaton can't silently drop the city from your resume. The font is Georgia, which ships with Windows and macOS. If Georgia isn't installed, as is often the case on Linux, Typst warns and falls back to Libertinus Serif.
Snippets so you don't memorize argument order
Four positional arguments are easy to get wrong, so VS Code snippets for atsresume and its Typst twin fill them in as tab stops: 16 snippets for atsresume (ats-experience, ats-education, ats-skill…) and 13 for readable-classic, the Typst package with the same layout (rc-experience, rc-skills…). ats-resume and rc-resume insert a whole file. For Typst you need the Tinymist extension, which provides the language mode they're registered for. And in Typst $ starts math, so a salary in a bullet has to be written \$25,000.
What came out of the PDFs
I rendered template.qmd with Quarto 1.10.19 and compiled the CTAN package's atsresume-example.tex with pdfLaTeX from TeX Live 2026. Both samples are the same made-up economics graduate, so the output is easy to compare. Then I extracted the text with pypdf 7.1.1 (page.extract_text(), default mode) and pdfminer.six 20260107 (extract_text() with default layout parameters).
Both PDFs were one page, and both extractors returned all 12 bullets as •. They also returned fi and fl as separate letters in "first", "finance" and "fluent". A literal keyword search can miss those words if extraction leaves a single fi or fl character in them.
pypdf put EDUCATION before the school in both files. Here's the start of its Quarto output:
Olivia Bennett
Providence, RI | (401) 555-0132 | olivia.bennett@example.com | linkedin.com/in/oliviabennett
EDUCATION
Halden College Providence, RI
Bachelor of Arts in Economics, Minor in Statistics Sep 2022 – May 2026
• GPA: 3.8/4.0, Economics Department Honors Program
In the pdfLaTeX file, though, pypdf ran two bullets together into one word each:
• Builtthree-statementmodelsfortwoconsumerretailclientsandupdatedcomparablecompanysetseachweek
In this pdfLaTeX build there are no space characters between words, so extractors have to infer word breaks from the gaps between letters. pdfminer.six got those two bullets right, and pypdf got them wrong. Both happen to be bullets that fill exactly one line. pdfTeX can write fake spaces (\pdfinterwordspaceon), but atsresume leaves that off on purpose: its source notes that OpenResume, a pdf.js-based resume parser, then reads every line as one long word. Compiling the same file with LuaLaTeX gave pypdf normal spaces in both bullets, and the Quarto PDF had no run-together words at all. For this sample, I'd compile with LuaLaTeX if pypdf might be the reader.
pdfminer.six had its own quirk. It groups text into blocks by position first, so the city and dates on the right margin became separate blocks. In the LaTeX PDF, EDUCATION came after the degree line. In the Quarto PDF, it landed in the middle of the coursework bullet, between its first line and the continuation "in Python". A parser that relies on those blocks could miss which section the school belongs to.
If you change the layout or switch LaTeX engines, run pypdf and pdfminer.six on the new PDF. Check whether EDUCATION comes before the school and whether the long bullets still have spaces.
When someone asks for a Word file
If an application asks for a .docx, look for Classic, the same layout as the Quarto format, among the LaTeX-style resume templates for Google Docs. Copy it, paste in your text and download it as a Word file.
I work on ReadableResume, which made these templates. The Quarto format and the snippets are MIT-licensed. atsresume is under the LaTeX Project Public License 1.3c.
I used AI to help draft this post.
Top comments (0)