IujinWee Garden 🌱

Resume Expert

4 min read

Introduction

This n8n agent aids users to refine resumes by fine-tuning resumes to fit the job description listed, based on the user’s past experience, skillsets & personality.

It also prepares a cover letter expressing the user’s job preference, which can increase the likelihood of passing the screening round.

Problem Statement

Applying for Job Interviews have always been time-consuming, especially reviewing job descriptions, technical competence & company cultural fit, etc.

Instead of manually having to review each job listing, this agent can automate the task of refining resumes to fit the job requirements, can drafting professional ready-made resumes which the user can directly use to send for job application.

Solution

This n8n workflow orchestrates this logic for refining the resume:

Entry & Data Acquisition

  1. User enters Telegram Bot: iujinwee-resume-expert

  2. User uploads URL of the job listing (i.e. LinkedIn URL)

  3. Extracts relevant information (A) from URL

    • Company-specific information (i.e. Title, Company, Culture)

    • Skillsets (i.e. Technical, Soft skills, Years of Experience)

Knowledge Retrieval

  1. Generate template for updating skills & experiences as KBA (Local storage)

    • Resume Template (i.e. Contact details, Structure)

    • Skills, Qualifications & Certifications

    • Past job experiences

  2. Automated KBA-refresh for quick retrieval (Optimization)

  3. Retrieves the following information (B) from KBAs (i.e. Google Drive, Local Storage)

    • To support image recognition

Intelligence & Synthesis

  1. Contexts (A) & (B) are sent to LLM to:

    • Generate high-impact, quantified resume pointers

    • Inject generated content into resume template

  2. Use LLM-as-judge for iterative resume refinement

    • Scoring metrics: Impactful, Relevancy, Keyword Density

    • Passing threshold: 80%

Delivery

  1. Upload final resume draft in DOCX to dedicated Google Drive folder.

  2. Send notification via Telegram Bot with shareable link.

Done! Your tailored resume is ready. 🚀Job Title: [Job Title] at [Company Name]Match Score: [Score]% (Based on your "Impact" metric)🔗 View Resume: [insert_gdrive_link] 📑 Original Listing: [insert_job_url]Good luck with the application!

Prompts (To be refined)

For our prompts, we’ll be using the XX framework which is effective for XX model.

Resume Expert

LLM Model:

Goal: Transform raw job requirements and your master history into tailored, high-impact bullet points.System Prompt: You are an expert Technical Career Consultant and Resume Strategist. Your goal is to synthesize Job Description data (Dataset A) with a User's Master History (Dataset B) to create a highly optimized resume draft.Operational Constraints:  1. Action-Oriented: Start every bullet point with a strong action verb (e.g., "Architected," "Orchestrated," "Streamlined").  2. Quantifiable Impact: Whenever possible, use the Google XYZ formula: "Accomplished [X] as measured by [Y], by doing [Z]."  3. Keyword Alignment: Naturally integrate technical skills and soft skills identified in Dataset A.  4. Conciseness: Keep pointers lean; focus on relevancy over volume.  5. Format: Output the content in structured JSON or Markdown as required by the template node.Task: Generate a list of resume pointers that bridge the gap between the user's past experiences and the specific culture and requirements of [Company Name].

LLM as Judge

LLM Model:

Goal: Act as a "gatekeeper" to ensure the resume meets a high professional standard before the user ever sees it.System Prompt: You are a Senior Technical Recruiter with a "no-nonsense" approach to evaluating candidates. You are auditing a tailored resume against a specific Job Description.Evaluation Rubric (Scoring 0-100):Relevancy (40%): Does the content directly address the specific tech stack and years of experience required?Impact (40%): Are achievements quantified with metrics, or are they just a list of "duties"?Clarity & Tone (20%): Is the language professional, free of fluff, and aligned with the company culture?Output Format (Strict JSON): { "score": integer, "passing": boolean (true if score >= 80), "critique": "Short explanation of why points were deducted", "suggestions": ["Specific improvement 1", "Specific improvement 2"] }Instruction: Be highly critical. If the resume feels generic or lacks data-driven results, score it below 80.

Architecture Diagram

Technologies

  • n8n workflow (i.e. Telegram, Google Drive)
  • KBAs (i.e. resume, skills & qualifications, resume template)
  • Database (i.e. metadata, KBAs)

Progress Tracker

  • Refine idea documentation
  • Prepare high-level architectural diagram
  • Agentic Development
    • Add n8n workflow
    • Add KBAs to local storage
    • Prompt fine-tuning using state-of-the-art frameworks
    • Add golden resume for reference
    • Store metadata in DB
  • Deploy services
    • n8n services
    • local LLM
  • Testing & Validation
  • Performance metrics … to see how effective this solution is!

Kudos for reaching this far! 🥂

This is my first agentic project with local LLM

Graph View