CV

Curriculum Vitae, built from a single structured data source.

How this works

The CV content lives in data/resume.json, a JSON Resume-based schema extended with a few custom sections (badges, talks, meta). Text fields use a small markup dialect — [text](url), **bold**, *italic*, `code` — defined in scripts/richtext.py.

scripts/build.py renders that data through the Jinja2 templates in templates/ to produce every output format from one source:

File	Purpose
cv_en.tex / cv_en.pdf	Human-facing LaTeX/PDF, same visual design as before
cv_en.md	Markdown, for GitHub rendering or pasting into other tools
llms.txt	Plain-text version, easy for LLMs/agents/ATS systems to parse
data/resume.json	The structured source of truth itself — also a valid machine-readable export

CI (.github/workflows/build-cv.yml) regenerates and commits all four on every push to main.

Editing the CV

Edit data/resume.json (not the generated .tex/.md/.txt files — those are overwritten by the build). Then run:

pip install -r scripts/requirements.txt
python3 scripts/build.py

This regenerates cv_en.tex, cv_en.md and llms.txt. Compiling cv_en.tex to a PDF requires a LaTeX distribution with lualatex (TeX Live with roboto, fontawesome, marvosym); CI does this automatically, so a local LaTeX install is only needed to preview the PDF before pushing:

latexmk -lualatex cv_en.tex

Editing templates

Changing the visual design or output layout means editing the templates in templates/, not data/resume.json:

• resume.tex.j2 — LaTeX template (uses \VAR{ } / \BLOCK{ } Jinja delimiters so they don't clash with LaTeX's own { } braces)
• resume.md.j2 — Markdown template
• resume.txt.j2 — plain-text/llms.txt template