My résumé, written in LaTeX and compiled in a reproducible Docker image — a clean single-page PDF that doubles as a small software project: versioned, built in CI, and layered with metadata that parsers, crawlers, and screen readers can read.

Why

It started with a PDF viewer showing "resume" instead of my name in its title bar. That small annoyance turned into a question — what is a PDF, what metadata can it carry, and how much of a résumé can a machine actually understand? The answer became this repo: a résumé treated like a production system rather than a Word file.

📝 Full write-up — Making my resume machine-readable.

Prerequisites

• Docker — builds the LaTeX toolchain
• Git

Optional, for verifying the output:

• exiftool — inspect embedded XMP metadata
• pdftotext (poppler) — confirm ATS text extraction
• python3 + pypdf — read embedded attachments

Note

Compilation runs inside a custom Docker image for consistency and reproducibility across environments. The image is amd64-only (pinned in the Dockerfile), so pass --platform=linux/amd64 on Apple Silicon and other arm64 hosts.

How to Use

1. Clone the repository

git clone https://github.com/anishshobithps/resume.git
cd resume

2. Build the Docker image

docker build -t latex-builder .docker

3. Compile the resume

docker run --rm --platform=linux/amd64 -v "$(pwd):/data" \
  latex-builder -jobname="Anish_Shobith_P_S_Resume" main.tex

Note

-jobname controls the output filename — change it as you wish. Run the command twice if you add cross-references or metadata that need a second pass to settle (the CI compiles twice for this reason).

ATS & accessibility

The PDF is engineered so automated systems read it as cleanly as a person does.

Vanity links that ATS still understand

The header displays short vanity URLs (n10nce.dev/github, n10nce.dev/linkedin) but tags each one with a PDF /ActualText entry through the \atslink macro (backed by the accsupp package). Text extractors, copy-paste, screen readers, and ATS parsers pull out the canonical github.com/… and linkedin.com/in/… strings — so profile-detection works — while the printed page stays branded and tidy. The link's href still points at the n10nce.dev redirect, keeping any click analytics intact.

Verify it:

pdftotext Anish_Shobith_P_S_Resume.pdf - | grep -iE 'github|linkedin'
# prints the canonical github.com / linkedin.com URLs, not n10nce.dev

Tagged for accessibility

formatting.sty enables Unicode glyph mapping (glyphtounicode, \pdfgentounicode) and marks the document as tagged, so extracted text is accurate and screen-reader friendly rather than a jumble of ligatures.

Embedded structured data

The compiled PDF carries rich metadata across several standards, readable by ATS platforms, semantic crawlers, and document parsers:

Standard	Source	Description
XMP / Dublin Core	\hypersetup in main.tex	Title, author, keywords, rights, language, dates
IPTC Core	\hypersetup in main.tex	Contact email, URL, address
Schema.org JSON-LD	schema.json (embedded)	Person, occupation, education, projects, skills with ATS aliases
JSON Resume	resume.json (embedded)	Open standard parsed by ATS such as Workday, Greenhouse, Lever

Note

Two traps the write-up documents (and this repo works around): \hypersetup must stay in main.tex — hyperxmp emits no XMP if it runs from a .sty file. And avoid LaTeX commands inside metadata values (\DTMtoday, \textendash, …); they silently corrupt the XMP block instead of erroring. Attachments use embedfile (spec-compliant name tree), not attachfile2, whose annotations parsers can't see.

Inspect the XMP metadata:

exiftool -xmp:all Anish_Shobith_P_S_Resume.pdf

Extract the embedded JSON attachments:

python3 -c "
import pypdf
r = pypdf.PdfReader('Anish_Shobith_P_S_Resume.pdf')
for name, data in r.attachments.items():
    print(f'--- {name} ---')
    print(data[0].decode('utf-8'))
"

Customization

• Content — edit the files in sections/; adjust PDF metadata in main.tex.
• Formatting — change layout, fonts, and macros in formatting.sty.
• Links — point sections/header.tex at your own handles; \atslink{href}{visible}{extracted} sets the displayed text and the canonical string ATS should read.
• Structured data — update schema.json and resume.json to reflect your own information.

Releases

Important

GitHub Actions rebuilds the image and publishes a new release on every push to main that touches the sources (and on manual dispatch).

Download the latest compiled PDF from the Releases page.

License

Licensed under the Apache-2.0 License. See LICENSE for details.