The Brutal AI: How We Built the Resume Roaster

Job hunting is undeniably one of the most stressful experiences in the modern professional world. What makes it even worse is the generic, sugarcoated advice you receive from friends and standard career counselors. "Just be yourself," they say. "Make sure to use action verbs."

Our PM Strategy Department realized that what job seekers actually need isn't polite encouragement; they need objective, harsh, and highly actionable feedback. They need to know exactly why their resume is being thrown into the digital trash bin by Applicant Tracking Systems (ATS).

To solve this, we decided to build the AI Resume Roaster—a tool designed to provide brutal, Silicon Valley-grade feedback for our users.

The Architecture: Why We Chose Claude 3 Opus

When building the engine, we evaluated several Large Language Models. We ultimately wired up Claude 3 Opus because it significantly outperformed competitors in deep contextual analysis and nuanced critique.

Our prompt engineering went through multiple rigorous iterations. Initially, the AI was far too polite. It would sandwich criticisms between layers of compliments, which diluted the impact. We had to aggressively enforce a "Gordon Ramsay of Tech" persona. The system prompt specifically instructs the LLM to output a precise, predictable JSON structure containing:

1. The Roast Score (0-100): A merciless numerical grade based on impact, clarity, and formatting.
2. Brutal Criticisms: No sugarcoating. If a bullet point is meaningless corporate jargon, the AI says so.
3. Actionable Fixes: It doesn't just tear you down; it provides a rewritten, high-impact version of your weak bullet points using the STAR (Situation, Task, Action, Result) method.

Handling PDF Parsing at the Edge

One of the biggest technical hurdles in this project was parsing PDFs cleanly on the server edge. Resumes come in wildly different formats—two-column layouts, heavy graphics, and weird fonts.

We utilized a customized version of pdf-parse running inside a lightweight serverless function. This allowed us to strip the raw text while maintaining the general hierarchy of the resume. If the text extraction fails (which happens with heavily flattened image-based PDFs), our frontend gracefully falls back to a raw text-input mode, ensuring the user is never stuck.

Why Brutal Feedback Works

During our beta testing, we learned something fascinating: users actually prefer harsh feedback when it is accurate and constructive. The Resume Roaster doesn't just critique; it rebuilds confidence by showing exactly how to fix the flaws.

It’s an incredibly resource-intensive tool, but seeing users completely transform their CVs and land interviews makes the server costs worth it. If you're currently job hunting and want to know why you aren't getting callbacks, you can try the Resume Roaster right now on our platform.

And just like we apply critical analysis to resumes, we apply the same brutal logic to startup ideas. If you are thinking of starting a business instead of getting a job, make sure to run your concept through our Idea Roaster to see if your SaaS idea is actually viable.

Don't settle for polite lies. Embrace the brutal truth and level up your career.