How the CV works

There is one file to edit: _data/cv_source.yml.

Everything else is generated from it and is overwritten on every build. Files carrying a DO NOT EDIT banner are generated — changes to them will be lost.

                        _data/cv_source.yml
                                 |
                     scripts/build_cv.py
                                 |
      +--------------+-----------+-----------+---------------+
      |              |           |           |               |
  _pages/cv.md  _publications/  _talks/  _teaching/     cv/cv.tex
   (CV page)     (/publications) (/talks) (/teaching)        |
                                                        files/cv.pdf
                                                     (download link)

Updating the CV

1. Edit _data/cv_source.yml.
2. Either run make cv locally, or just commit and push — the Build CV GitHub Action regenerates everything and commits the result back.

Adding a paper means adding one block to the publications: list. That single block produces the CV entry, the /publications/ page entry, and the line in the PDF. There is nowhere else to update.

Commands

make site works with only Python + PyYAML (make deps installs it). make pdf additionally needs LaTeX. On macOS the lightweight option is TinyTeX:

curl -sL https://yihui.org/tinytex/install-bin-unix.sh | sh
tlmgr install moderncv fontawesome5 xurl enumitem latexmk

If LaTeX is missing or incomplete, make cv no longer fails: it regenerates all the site files, prints a SKIPPED: line explaining what is absent, and exits cleanly. make texdeps then diagnoses the toolchain and prints the exact command to fix it.

A common macOS wrinkle: BasicTeX installs to a root-owned tree, so tlmgr needs sudo. And if tlmgr reports “Local TeX Live (N) is older than remote repository (N+1)”, your distribution is a full release behind CTAN and cannot install anything until it crosses releases:

curl -L https://mirror.ctan.org/systems/texlive/tlnet/update-tlmgr-latest.sh -o /tmp/upd.sh
sudo sh /tmp/upd.sh --update
sudo tlmgr update --self --all
sudo tlmgr install moderncv fontawesome5 xurl enumitem

If you would rather not install LaTeX at all, skip it — push and let the Action build the PDF. make site still works fine without it.

The YAML in brief

Sections and their order come from section_order: at the top. Reorder or delete entries there and both the website and the PDF follow. No code changes.

Publications — each entry takes:

  - slug: short-name-for-the-url
    date: 2025-02-01        # required by Jekyll; controls ordering
    year: 2025              # shown in the CV
    authorship: first       # first | co  -> which CV subsection
    authors: Hoffman, L., F. Massonnet, and A. Sticker.
    title: Full title of the paper.
    venue: Journal Name
    details: 2, e2025JH000669
    doi: 10.1029/2025JH000669
    excerpt: One or two sentences shown on the publication page.

Optional: status: in_progress (renders as “In progress” and, with site: false, is kept off the website until it is published).

Talks — two distinct fields:

  - slug: 2025-esip-machine-learning-cluster
    date: 2025-01-01
    year: 2025
    group: invited                # invited | contributed -> which CV section
    format: Oral presentation     # Oral presentation | Poster | Seminar
    title: Explainable AI for Arctic sea ice prediction   # the TALK's title
    meeting: Earth Science Information Partners (ESIP) Machine Learning Cluster Meeting
    venue: Earth Science Information Partners             # host organisation
    location: Virtual

title: is the talk’s own title and meeting: is the event it was given at. The CV line renders as:

2025: “Explainable AI for Arctic sea ice prediction.” Earth Science Information Partners (ESIP) Machine Learning Cluster Meeting. Virtual. Oral presentation.

Entries still carrying title: TODO fall back to the meeting name alone, so the site and PDF stay correct while you fill them in — and each build reports how many are left. If a talk genuinely had no distinct title, delete its title: line entirely and it renders from meeting: permanently, with no warning.

Teaching — teaching: holds courses, guest_lectures: holds one-off lectures. Both feed the /teaching/ collection.

Talks split across sections by their group: value, mapped near the top of the YAML:

talk_sections:
  invited_talks: invited
  presentations: presentation
  meetings_workshops: meeting

Change a talk’s group: to move it between sections; add a line here (and to section_order:) to create a new one.

Compact sections

Mentorship, leadership, service and teaching use a one-line-per-entry style:

mentorship:
  compact: true
  entries:
    - dates: Fall 2024–Spring 2025
      who: Antonio Martinez Soares
      role: Master's Student
      org: Université catholique de Louvain

renders as Fall 2024–Spring 2025 | Antonio Martinez Soares, Master's Student | Université catholique de Louvain. The line is assembled from whichever of who, role, org, note and text are present, so you can use as few as you like. Set compact: false on a section to go back to bulleted entries.

Teaching entries additionally take a summary: — the one-line version shown on the CV — while keeping their bullets: for the fuller /teaching/ page.

Bold position titles

In compact sections the role: field is rendered in bold — that is the only part that bolds, so keep the position title there and put everything else in text::

    - dates: 2025–Present
      role: Co-Lead                       # <- bold
      text: OAISIS Working Group; ...     # <- plain

Teaching entries carry both: role: is the bolded position on the CV, while title: (e.g. “Instructor, Climate Change and the Ocean”) still names the /teaching/ collection page. Guest lectures default to Guest Lecturer unless you set role: yourself.

Turn it off globally with bold_roles: false.

Bold author name

Your name is emboldened automatically in every publication author list. The default matches Hoffman, L. and L. Hoffman; override with:

basics:
  bold_name:
    - Hoffman, L.
    - L. Hoffman

Adding a brand-new section

Two steps, both in _data/cv_source.yml:

1. Add the content under a new top-level key.
2. Add that key to section_order: — this is the step that makes it appear. A section not listed there is not rendered anywhere. The build prints a warning if you forget.

No Python changes are needed. An unrecognised key is rendered generically, in any of these shapes:

personal_interests:                       # list with optional hint labels
  - text: Bouldering, backpacking, guitar, violin, yoga, and cats.

personal_interests:                       # or plain bullets
  - Bouldering and backpacking
  - Guitar and violin

personal_interests: A sentence of prose.  # or a single paragraph

The hint column (the left-hand label in the PDF) is picked up from any of date, dates, label, term, or year on an entry — the same shape the honors: and activities: sections use.

The heading is derived from the key: personal_interests becomes “Personal interests” on the site and “Personal Interests” in the PDF. To choose your own wording, add an override:

section_titles:
  personal_interests: Outside the Lab

Why pipes are written as |

Your _config.yml runs kramdown with input: GFM. That parser treats any line containing a | as a table — it does not require a header separator row. A literal pipe in the CV therefore renders as a stray bordered table on the website.

The generator writes | in markdown instead. Table detection scans raw text and sees no pipe, while entity_output: as_char turns the entity back into a real | in the final HTML — so the page shows the pipe you want with no table. The PDF is unaffected and uses \textbar{}.

This happens automatically for every field, so you can type ordinary pipes in cv_source.yml. A build-time guard warns if a raw pipe ever reaches _pages/cv.md.

Emoji and symbols

The website renders any Unicode you like. pdflatex cannot typeset most symbols and emoji, so the build silently drops them from the PDF and prints a note telling you which characters it removed. Accented Latin characters (é, è, ç, ü …) are fine in both.

If you want a symbol in the PDF badly enough, switch the engine to XeLaTeX by adding $pdflatex = 'xelatex %O %S'; to a cv/.latexmkrc and using a font with the glyph — but dropping it is usually the easier answer.

Styling the PDF

cv/preamble.tex holds the LaTeX preamble — document class, moderncv style and colour, margins, packages. It is created once and then belongs to you: the generator never overwrites it. Change \moderncvstyle{classic} to banking, casual, or oldstyle, or \moderncvcolor{blue} to another colour, and rebuild.

cv/cv.tex is the generated body. Do not edit it.

Caveat: talk and publication dates

Your source CV recorded years only, but Jekyll requires a full YYYY-MM-DD for every collection entry. Where the real month was unknown, dates are placeholders in January, with the day used to preserve your CV’s ordering within each year. They are correct in ordering but not in month. Replace them with real dates in cv_source.yml as convenient — the day-level ordering trick stays valid either way, since sorting is newest-date-first.

If a generated file goes missing

.cv-generated.json records what the last build produced. Anything in that list that a later build no longer produces gets deleted automatically — that is how renaming a slug cleans up its old file. Delete the manifest if it ever gets out of sync; the next build recreates it.