the system · 03 · application engine

How my applications happen.

Applying used to mean copy-pasting a CV and hoping it’s current. I turned it into a pipeline instead, the same discipline I use for deploying software, applied to applying for work.

> “Apply to this role at …”
✓ position captured, company, role, posting
✓ stations selected · wording tailored
✓ cv.pdf + cover-letter.pdf typeset
✓ draft staged in my mailbox, I hit send
# one sentence in, a finished application out
the pipeline
  1. 01

    One content database

    Every fact about me, roles, projects, bullets, lives once, versioned and bilingual. No document ever goes stale.

  2. 02

    A position arrives

    A job posting is captured as a position: company, role, the full text, the target everything else adjusts to.

  3. 03

    Wording, tailored

    AI tunes emphasis and selection for that position, which stations matter, which bullets lead. It can rephrase; it can never invent a fact.

  4. 04

    Typeset, not exported

    CV and cover letter render through LaTeX templates into clean, deterministic PDFs, every version stored.

  5. 05

    Tracked to the end

    Position, documents and stage live in one pipeline, from first draft to offer, nothing in loose folders.

same source of truth

The database behind my CVs is the same one rendering this website. Change a fact once, it’s current everywhere.

want the deep dive? · paper
Facts Before Phrasing

I wrote this pipeline up as a paper, the fact-vs-phrasing model, the architecture, and a blinded evaluation where generating from the fact base cut fabrication from 78% to zero. System name: Dossier.

next · 01
Homelab
the machines