AI Gatekeeper: A Highly Efficient Rejection-as-a-Service Platform 😂

#ai #machinelearning #devchallenge #googleaichallenge

This is a submission for the DEV April Fools Challenge

What We Built
AI Gatekeeper is a gloriously over-engineered rejection machine. You can submit code, a resume, a cover letter, or just a sentence you regret in hindsight, and the app will process it through three escalating rounds of corporate disappointment.

It looks like a serious candidate evaluation system, but the entire point is that it behaves like one of those polished enterprise tools that somehow makes you feel less employable after every click. The UI starts with confident analytics and ends with a final verdict that should probably be reviewed by legal.

Demo

Demo
Live app: https://aigatekeeper.vercel.app/

Code

The project is built as a tiny Vercel app with exactly what it needs and nothing it deserves.
Github : https://github.com/Vk2245/DEV-April-Fools-Challenge

Frontend: pure HTML, CSS, and vanilla JavaScript in index.html
Backend: Vercel serverless function in evaluate.js
Routing: vercel.json
Environment example: .env.example

How I Built It

The frontend is a single-page experience designed to feel like an enterprise talent platform. It uses custom CSS variables, a responsive layout, and staged UI states so each round feels more alarming than the last.

The backend accepts a message history and a round number, builds a round-aware system prompt, and sends the request to an LLM. Gemini is used first, with Groq as fallback if the primary provider fails. The response is validated before it reaches the UI, because even rejection needs standards.

Tech stack:

Vanilla HTML, CSS, JavaScript
Vercel Serverless Functions
Gemini API
Groq API fallback
Local .env support for development

Prize Category

Best Google AI Usage
This project uses Google Gemini as the primary model behind the entire evaluation flow. The backend builds a round-specific system prompt, passes the submission history through the API, and turns Gemini into the engine for the three-stage rejection pipeline. The AI is not a side feature here; it is the entire point, and unfortunately for the user, it is very committed to its job.

What Makes It Funny

The joke is not just that the app rejects everything. The joke is that it rejects everything with the tone of a polished product review, as if your resume and your code were both entered into a committee meeting and unanimously described as “not strategically aligned.”