Resume Pipeline

A Markdown-to-PDF resume system with AI-assisted tailoring for job applications.

---

Why this exists

Tailoring a resume for every application by hand is slow, error-prone, and produces drift across versions. LaTeX gives great output but is hostile to quick edits. This repo solves both problems: Markdown as the single editable source, Pandoc + a Lua filter to render it into a pixel-perfect PDF, and an AI agent that tailors the whole thing to a job description in one shot.

flowchart LR
    JD[Job Description] --> Agent
    Agent --> MD[tailored resume.md]
    MD --> Pandoc
    Pandoc --> |Lua filter| LaTeX[LaTeX AST]
    LaTeX --> |pdflatex| PDF[Tushar-Pandey-resume.pdf]
    PDF --> local[applications/role-NNN/]

Loading

The agent reorders bullets, rewrites the summary, and adjusts the skills section — without fabricating anything not already in the base resume. Company names never appear in the published output.

---

How it works

Pipeline

flowchart TD
    src["src/resume.md\n(Markdown source)"]
    tmpl["templates/resume-template.tex\n(LaTeX preamble + $body$)"]
    lua["filters/resume.lua\n(Pandoc Lua filter)"]
    pandoc["pandoc --pdf-engine=pdflatex"]
    pdf["Tushar-Pandey-resume.pdf"]

    src --> pandoc
    tmpl --> pandoc
    lua --> pandoc
    pandoc --> pdf

Loading

src/resume.md — the only file you edit day-to-day. Plain Markdown with structural conventions the Lua filter understands.

filters/resume.lua — translates the Markdown AST into custom LaTeX resume commands:

Markdown pattern	LaTeX output
H1 + contact paragraph	\begin{center} heading block
H2	\section{}
H3 + italic subline	\resumeSubheading{company}{date}{role}{location}
H3 **bold** | _italic_	\resumeProjectHeading{\textbf{…} $|$ \emph{…}}{}
Bullet list	\resumeItemListStart / \resumeItem{} / \resumeItemListEnd

templates/resume-template.tex — full LaTeX preamble (packages, margins, custom commands). Body is $body$.

config/build.conf — pipeline configuration, including the output PDF filename:

OUTPUT_FILENAME="Tushar-Pandey-resume.pdf"

Markdown conventions

# Full Name
phone | email | [linkedin](url) | [github](url)

## Experience

### Role Title | Company Name | Date Range
_Team description | City, Country_

- **Achievement**: what you built and why it mattered.

## Key Projects

### **Project Name** | _Tech Stack_

- What it does.

## Technical Skills

- **Category**: item, item, item

The only non-obvious rule: the last | in an H3 line is the date separator, so company names containing | work without escaping.

Building

# Prerequisites
brew install pandoc
# BasicTeX: https://tug.org/mactex/morepackages.html

# Build base resume
./scripts/build-resume.sh

# Build a tailored variant
./scripts/build-resume.sh \
  applications/<role-type>/<company-slug>/resume.md \
  applications/<role-type>/<company-slug>/

Output filename is set in config/build.conf. Default: Tushar-Pandey-resume.pdf.

---

AI-assisted tailoring

sequenceDiagram
    participant You
    participant Agent
    participant Pipeline

    You->>Agent: paste job description
    Agent->>Agent: identify role-type, company slug
    Agent->>Agent: tailor resume.md from src/resume.md
    Agent->>Pipeline: build-resume.sh applications/…/resume.md
    Pipeline-->>Agent: PDF (saved locally in applications/)
    Agent->>Agent: update APPLICATIONS.md
    Agent->>You: PDF ready locally, comms drafted

Loading

The agent:

• Rewrites the profile summary to speak to the role
• Reorders bullets within each job to front-load what the JD values
• Leads the skills section with the role's key technologies
• Pulls extra detail from base/Tushar-Pandey-resume.md when the base src/ is lighter

Hard constraint: no fabrication. Every metric, technology, and bullet in the output must exist in the base resume.

Privacy model

Everything stays local. Nothing is pushed to a remote repository.

• applications/ — tailored resumes, compiled PDFs, per-application artifacts (company names, JDs, comms)
• APPLICATIONS.md — local index mapping IDs to companies and tracking status

Company names, job descriptions, and all application data exist only on your machine.

---

Repo structure

├── src/resume.md                  # Pandoc Markdown source (edit this)
├── base/                          # Full content library + original LaTeX (do not modify)
├── templates/resume-template.tex  # Pandoc LaTeX template
├── filters/resume.lua             # Lua filter (AST → custom LaTeX commands)
├── config/build.conf              # Pipeline config (output filename, etc.)
├── scripts/build-resume.sh        # Build script
├── applications/                  # Tailored sources, compiled PDFs, per-application artifacts
│   └── <role-type>/<company-slug>/
├── .raven/                        # Agent workflow artifacts
└── compile.sh                     # Legacy pdflatex script (still functional)

---

See also

• AGENTS.md — agent instructions: tailoring workflow, directory model, Markdown conventions
• CONTRIBUTING.md — human guide: editing, building, index format