I built a Resume Interview Bot because ChatGPT was too generic

#ai #webdev #javascript

I've been prepping for engineering interviews lately, and I noticed a problem.

If I ask ChatGPT: "Roast my resume," it gives generic advice like "Add more metrics."
If I ask: "Simulate a technical interview," it asks standard questions like "What is the difference between TCP and UDP?"

The missing link was Context. The LLM didn't know my specific history.

So, I built a tool to fix it.

The Project: Invoxy

I built Invoxy.xyz. It’s a free tool that parses your PDF resume and generates context-aware interview questions.

How it works under the hood

Client-Side Parsing: I used pdf-parse to extract raw text strings from the resume upload directly in the browser (keeping data private).
Context Injection: I built a system prompt that feeds the LLM your specific bullet points + the target Job Description.
The "Red Team" Persona: I tuned the model to be skeptical. Instead of being nice, it specifically looks for gaps in employment or vague claims (e.g., "You claimed 20% growth here—how exactly did you attribute that to your code?").