RoastMyResume

Paste your resume. Get roasted. Get better. An AI-powered savage critique tool that finds every weak line and shows you how to fix it.

Comment

Inspiration

I was job hunting and kept getting conflicting resume advice. Career coaches charge $200+ per session. Friends say "looks great" without reading it. AI tools give generic feedback like "add more metrics." I wanted something brutally honest—an app that would roast my resume line by line and actually tell me what's wrong. So I built it.

What It Does

RoastMyResume is an AI-powered savage resume critique tool. You paste your resume text, click "Roast It," and the app:

  • Calculates a Roast Score out of 100 based on issue severity
  • Identifies weak lines: passive voice, buzzwords, missing metrics, generic responsibilities
  • Delivers a unique, savage burn for each problematic line
  • Offers "Fix It" mode that rewrites weak bullets into recruiter-friendly language

No login. No paywall. Just paste and get roasted.

How I Built It

I built the entire full-stack application using MeDo—describing features in natural language and letting the AI generate the code. The tech stack MeDo assembled includes:

  • Frontend: React with dark-themed UI and orange accent design
  • Backend: Node.js with AI processing pipeline
  • AI/LLM: Integrated via MeDo's native AI capabilities for resume analysis and critique generation
  • Deployment: One-click deploy to public URL via MeDo's platform

Build Process

Step 1 — Core Scaffold (1 prompt): I described the entire app concept in a single prompt: text input, roast engine, score display, dark header with orange accent. MeDo generated the full-stack application with working UI and AI integration.

Step 2 — Deepening the Roast Engine (1 prompt): I asked MeDo to add severity-weighted analysis—detecting passive voice, buzzword density, unquantified impact, and generic responsibilities. Each issue type got a different deduction weight (Critical: -15, Medium: -10, Minor: -5).

Step 3 — Fix It Mode (1 prompt): I added a second button that transforms critique cards into side-by-side rewrites. The AI replaces passive voice with action verbs, adds placeholder metrics, and removes buzzwords while preserving original meaning.

Step 4 — Bug Fixes & Edge Cases (3 prompts): The hardest challenge was teaching the AI to understand resume structure. Initially, it roasted names ("MOUNESH KODANGAL"), section headers ("Technical Skills"), and contact information as if they were achievement bullets. I had to add a pre-processing step that categorizes each line before applying roast logic. Multi-line bullet points were also being split into separate cards—fixed by detecting sentence continuations.

Challenges I Faced

Resume Structure Recognition: The AI initially couldn't distinguish between a person's name and a project name, or between a section header and an achievement bullet. This required multiple iterations of categorization logic—teaching the system to identify [Name/Contact], [Section Header], [Skills List], and [Achievement Bullet] as distinct types with different roasting rules.

Template Repetition: The AI defaulted to a small set of generic burns ("Numbers, ever heard of them?") rather than crafting unique critiques for each line. I learned that AI needs explicit constraints to avoid template-based outputs.

Multi-Line Bullet Combining: Long achievement bullets that wrapped across multiple lines were being split into separate roast cards. Fixed by adding continuation detection logic.

Credit Management: Building within MeDo's credit system taught me to be precise with prompts. Every iteration cost credits, so I learned to structure fix requests with clear root cause analysis, fix instructions, and acceptance criteria.

What I Learned

  • Prompt Engineering: The difference between a working app and a broken one often came down to how precisely I described the expected behavior. Vague prompts produced vague features.
  • AI Limitations: LLMs default to templates when asked to be creative. Breaking that pattern required explicit anti-template constraints.
  • Resume Structure Matters: Building this app forced me to think deeply about what makes a resume line strong versus weak—knowledge I'm now applying to my own job search.
  • Ship Early, Fix Later: The first version roasted names and headers. Rather than restarting, I iterated. Perfect is the enemy of deployed.

What's Next for RoastMyResume

  • File upload support — accept PDF/DOCX files directly
  • Roast intensity slider — from "Gentle Nudge" to "Career Eulogy"
  • Comparison mode — roast two versions of the same resume to see which is better
  • Industry-specific roasts — tech resumes get different burns than consulting resumes

Built With

Share this project:

Updates