Resume as Code

This project implements a "Resume as Code" philosophy, leveraging LLMs (Large Language Models) and structured data to automate the creation of highly tailored, professional resumes.

It solves the common pain points of resume maintenance:

• Fragmentation: Keeping multiple versions of Word/PDF files for different job applications.
• Inconsistency: Difficulty in maintaining consistent formatting and content across versions.
• Time-Consuming: Manually tailoring resumes for each Job Description (JD) is tedious.

By maintaining a single "Master Timeline" of your career and using AI agents to dynamically assemble resumes, you can generate a perfect match for any job opportunity in minutes.

View Full Resume PDF (EN)

查看完整中文版简历 PDF (ZH)

✨ Core Features

• Single Source of Truth: All career data (work, projects, education, certificates) is stored in modular YAML files.
• Resume Builder App: A local-first workbench with live YAML editor, real-time preview, deterministic pagination, and one-click PDF export.
• AI-Powered Agents:
  • Resume Generation Agent: Analyzes JDs and assembles targeted resumes.
  • Timeline Polishing Agent: Polishes raw experience descriptions using STAR/3W methodologies.
  • Interview Preparation Agent: Generates comprehensive interview guides based on your resume and target JD.
• Dual YAML Format Support: Supports both the legacy YAMLResume format and a new richer custom schema (v1.0), with automatic detection and seamless compatibility.
• Automated Validation: Integrated validation ensures generated resumes are syntactically correct and ready for rendering.
• Multi-Format Export: Supports exporting to high-fidelity PDF via the built-in Puppeteer export service, plus HTML and Markdown via the YAMLResume compiler.
• Multi-Language Support: Supports English, Simplified Chinese, Traditional Chinese (HK/TW), Spanish, French, and Norwegian, with a one-click language toggle in the builder.

---

🖥️ Resume Builder App

The resume-builder-app/ directory contains a local-first resume workbench built with React, Vite, Monaco Editor, and Puppeteer.

Features

Feature	Description
Split Editor + Preview	Monaco YAML editor on the left, pixel-perfect A4 preview on the right
Live Validation	Real-time YAML syntax and schema validation with inline diagnostics
Deterministic Pagination	Fine-grained block-level pagination with proper keep-together logic
Optimize Layout	One-click spacing optimization with adjustable scale (0.7×–1.3×)
PDF Export	Local Puppeteer + headless Chromium service for high-fidelity PDF generation
Language Toggle	Switch between 中文 and English rendering with one click
Preview Zoom	Floating zoom control (scale only — does not affect rendering/export)
Dual Format Support	Auto-detects legacy yamlresume or new schema v1.0 YAML
Section Ordering	Respects layouts[].sections.order (legacy) or top-level order (new schema)

Quick Start

cd resume-builder-app

# Install dependencies (also installs Chrome for Puppeteer)
pnpm install

# Start dev server (web preview + PDF export service)
pnpm dev
# → Web UI: http://localhost:5173
# → PDF Export Service: http://localhost:3001

# Production build
pnpm build
pnpm start

Architecture

resume-builder-app/
├── src/
│   ├── app/
│   │   ├── App.tsx              # Workspace shell (editor + preview + toolbar)
│   │   ├── renderer/
│   │   │   ├── ResumeRenderer.tsx    # Converts RenderModel → fine-grained blocks
│   │   │   ├── PaginatedPaper.tsx    # Measures blocks → A4 page containers
│   │   │   └── PrintStyles.tsx       # Print/PDF CSS overrides
│   │   └── routes/
│   │       └── PrintRoute.tsx        # Headless print route for PDF export
│   ├── compiler/
│   │   ├── compile-legacy.ts         # yamlresume → RenderModel adapter
│   │   └── compile-new-schema.ts     # New schema v1.0 → RenderModel compiler
│   ├── schema/
│   │   └── resume-schema.ts          # TypeScript types for new YAML schema v1.0
│   ├── models/
│   │   └── index.ts                  # Normalized RenderModel (single renderer input)
│   └── export-service/
│       └── server.ts                 # Express + Puppeteer PDF export service
├── package.json
└── vite.config.ts

---

📋 YAML Resume Formats

The Resume Builder supports two YAML formats. Files are auto-detected at load time.

New Schema (v1.0) — Recommended

The new format is a richer, versioned schema owned by the builder app. It supports discriminated section types, stable IDs, visibility flags, layout hints, and flexible content structures.

schema:
  version: '1.0'
  generator: resume-builder-app

document:
  title: 'My Resume'
  language: en # en | zh-hans | zh-hant-hk | zh-hant-tw | es | fr | no

basics:
  name: Zhang Wei
  headline: 'Senior Data Engineer · 7 YOE'
  phone: '+86 138 8888 8888'
  email: zhangwei@outlook.com
  url: zhangwei.dev
  summary:
    - 'Bullet point one describing core competency.'
    - 'Bullet point two with specific achievements.'
  profiles:
    - network: GitHub
      url: https://github.com/zhangwei
      username: github.com/zhangwei

# profiles can also be placed here (sibling of basics) for yamlresume compatibility
# profiles:
#   - network: GitHub
#     url: https://github.com/zhangwei
#     username: github.com/zhangwei

sections:
  - id: work
    type: work
    title: 'Work Experience'
    items:
      - id: job-1
        name: Acme Corp
        position: Senior Engineer
        startDate: '2022-01'
        endDate: '2024-06'
        summary:
          - 'Led migration of data platform to lakehouse architecture.'
          - 'Reduced pipeline latency by 60% through stream processing.'
        keywords: [Spark, Kafka, AWS, Terraform]

  - id: education
    type: education
    title: 'Education'
    items:
      - institution: MIT
        degree: Master
        area: Computer Science
        startDate: '2018-09'
        endDate: '2020-06'

  - id: projects
    type: projects
    title: 'Projects'
    items:
      - id: proj-1
        name: Data Platform v2
        description: 'Enterprise-scale lakehouse platform'
        startDate: '2023-03'
        endDate: '2024-01'
        summary:
          - 'Designed and implemented end-to-end lakehouse on AWS.'
        keywords: [Databricks, Delta Lake, Airflow]

  - id: skills
    type: skills
    title: 'Skills'
    items:
      - name: Languages
        keywords: [Python, SQL, TypeScript, Go]
      - name: Data
        keywords: [Spark, Kafka, Airflow, dbt]

  - id: certificates
    type: certificates
    title: 'Certificates'
    items:
      - name: AWS Certified Data Analytics
        issuer: Amazon Web Services
        date: '2023-05'

  - id: langAndInterests
    type: langAndInterests
    title: 'Languages & Interests'
    languages:
      - language: English
        fluency: Professional Working
      - language: Mandarin Chinese
        fluency: Native
    interests:
      - name: Open Source
        keywords: [data tooling, contributor]

# Optional: explicit section display order
order:
  - work
  - education
  - projects
  - skills
  - certificates
  - langAndInterests

layout:
  template: jake
  page:
    margins: { top: 1.5cm, left: 1.5cm, right: 1.5cm, bottom: 1.5cm }
    showPageNumbers: true
  typography:
    fontSize: 11pt

themeOverrides:
  spacing:
    sectionGap: 12
    entryGap: 8

Legacy Format (YAMLResume)

The YAMLResume format uses a flat content object with fixed section keys. The builder auto-detects this format by the presence of content at the root.

locale:
  language: zh-hans

layouts:
  - engine: latex
    sections:
      aliases:
        basics: 基本信息
        work: 工作经历
        education: 教育经历
        projects: 项目经历
        skills: 专业技能
        certificates: 证书
      order:
        - basics
        - education
        - work
        - projects
        - skills
        - certificates
    page:
      margins: { top: 1.5cm, left: 1.5cm, right: 1.5cm, bottom: 1.5cm }
      showPageNumbers: true
    template: jake
    typography:
      fontSize: 11pt

content:
  basics:
    name: John Doe
    phone: '+1 555-123-4567'
    email: john.doe@example.com
    url: https://johndoe.dev
    summary: |
      - 7 years of data engineering experience with Python and SQL.
      - Experienced in ETL automation, data quality, and monitoring.
  profiles:
    - network: GitHub
      url: https://github.com/johndoe
      username: johndoe
  education:
    - institution: Example University
      degree: Bachelor
      area: Computer Science
      startDate: Sep 1, 2015
      endDate: Jul 1, 2019
  work:
    - name: Acme Inc
      position: Senior Engineer
      startDate: Nov 14, 2022
      endDate: Now
      summary: |
        - Built enterprise data platform from scratch.
      keywords: [Python, React, AWS]
  projects:
    - name: Data Platform
      description: Enterprise lakehouse
      startDate: Mar 1, 2023
      endDate: Jan 1, 2024
      summary: |
        - End-to-end lakehouse on AWS.
      keywords: [Databricks, Airflow]
  skills:
    - name: Languages
      keywords: [Python, SQL, TypeScript]
  certificates:
    - name: AWS Certified
      issuer: AWS
      date: May 2023

Format Differences

Feature	Legacy (YAMLResume)	New Schema (v1.0)
Root structure	content object with fixed keys	basics + sections array
Section definition	Fixed keys (work, education, etc.)	Discriminated union with type field
Section ordering	layouts[].sections.order	Top-level order array
Section titles	layouts[].sections.aliases	Per-section title field
Item IDs	Not supported	Optional id on each item
Visibility control	Not supported	visible flag per section/item
Summary format	Multiline pipe-string (|)	Array of strings or pipe-string
Profiles location	content.profiles (sibling of basics)	basics.profiles or root profiles
Language setting	locale.language	document.language
Layout config	layouts[] array (multi-engine)	Single layout object
Theme overrides	Not supported	themeOverrides (colors, spacing)
Schema versioning	Not supported	schema.version field

---

🏗️ Architecture & Workflow

The system operates through three primary AI agents:

1. Timeline Polishing Agent

Input: Raw text description of a job or project. Output: Structured, polished YAML file in timelines/gem/.

1. Input Analysis: Identifies if the input is Work Experience or a Project.
2. Polishing: Applies STAR (Situation, Task, Action, Result) for work or 3W (What, Why, How) for projects.
3. Enrichment: Infers relevant technical keywords and industry context.
4. Storage: Saves the polished artifact to the timeline library.

2. Resume Generation Agent

Input: Target Job Description (JD). Output: A complete, tailored resume YAML file in resumes/gem/.

1. Job Analysis: Extracts key skills, requirements, and role context from the JD.
2. Matching: Selects the most relevant experiences from the Timeline library based on the analysis.
3. Section Generation: Generates tailored Summary, Skills, Work, and Project sections.
4. Assembly: Combines all sections with static profile data (Education, Certificates) into a final YAML file.
5. Validation: Validates the output against the schema.

3. Interview Preparation Agent

Input: Resume, JD Analysis, Company Business Analysis. Output: A comprehensive Interview Preparation Guide in interviews/gem/.

1. Input Verification: Ensures all necessary context files are present.
2. Strategy Generation: Creates a personal introduction strategy tailored to the role.
3. Deep Dive: Generates STAR-based deep dive questions for every project.
4. Q&A Bank: Creates an extensive technical Q&A bank covering specific tech, architecture, and domain knowledge.
5. Behavioral & Reverse: Prepares behavioral questions and high-quality reverse interview questions.

📂 Directory Structure

.
├── resume-builder-app/  # Local-first resume workbench (see above)
├── profiles/            # Personal basic information (static)
├── resumes/             # Resume generation artifacts
│   ├── gem/             # Final generated resume files
│   ├── temp/            # Intermediate generation artifacts
│   ├── job-analysis-prompt.md
│   ├── company-business-analysis-prompt.md
│   └── section-*-prompt.md
├── timelines/           # Master Timeline Library
│   ├── gem/             # Polished timeline event files (YAML)
│   ├── timeline-project-prompt.md
│   └── timeline-work-experience-prompt.md
└── interviews/          # Interview preparation artifacts
    ├── gem/             # Generated interview guides
    └── interview-prompt.md

🤖 AI Tools Support & Configuration

This project leverages AI agents to automate the resume generation process. While it can work with various LLMs, it is optimized for the following tools:

1. GitHub Copilot

• Status: Supported (Recommended)
• Configuration: No specific configuration is required. You can directly interact with Copilot Chat in VS Code.

2. Trae

• Status: Supported
• Configuration:
  1. Create Agent: Create a new Custom Agent in Trae.
  2. Configure Rules: Copy the content of .trae/rules/project_rules.md and paste it into the agent's instructions.
  3. Save: Name the agent resume-as-code.
• Usage: Select the resume-as-code agent when working on this project to ensure strict adherence to the workflow.

🚀 Usage Guide

Prerequisites

• Node.js (v18+) & pnpm installed.
• An LLM interface (e.g., GitHub Copilot Chat in VS Code).

Step 1: Build Your Timeline Library

Don't write a resume yet. First, build your database of experiences.

1. Open Copilot Chat.
2. Paste a raw description of a past job or project.
3. The Timeline Polishing Agent will format it into a structured YAML file in timelines/gem/.
4. Review and save the file.

Step 2: Configure Static Data

Fill in your static information in the profiles/ directory:

• profiles/basics.yml: Contact info, social links.
• profiles/education.yml: Academic history.
• profiles/certificates.yml: Certifications.

Step 3: Generate a Resume

When you find a job you want to apply for:

1. Copy the Job Description (JD).
2. Paste it into Copilot Chat.
3. The Resume Generation Agent will:
   • Analyze the JD.
   • Select relevant timeline events.
   • Generate tailored content.
   • Assemble a final YAML file in resumes/gem/ (e.g., Name_JobTitle_Company.yml).

Step 4: Preview & Export PDF

Use the Resume Builder app to preview and export:

cd resume-builder-app
pnpm install
pnpm dev

1. Open http://localhost:5173 in your browser.
2. Load your generated YAML file in the editor (or use sample loading).
3. Preview the formatted resume with live pagination.
4. Use ⚡ Optimize Layout to fine-tune spacing if needed.
5. Click Export PDF for a high-fidelity PDF download.

Alternatively, use the YAMLResume CLI for LaTeX-based PDF:

pnpm yamlresume build "resumes/gem/Your_Resume.yml"

Step 5: Prepare for Interview

Once your resume is ready:

1. Provide the generated Resume, JD Analysis, and Company Business Analysis to Copilot Chat.
2. The Interview Preparation Agent will generate a detailed guide in interviews/gem/.
3. Use this guide to practice your introduction, project deep dives, and technical Q&A.

🛠️ Development

# Install dependencies (root)
pnpm install

# Resume Builder App
cd resume-builder-app
pnpm install
pnpm dev          # Start dev server + PDF export service
pnpm build        # Production build
pnpm typecheck    # Type checking

📄 License

MIT