<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Forem: VISHAL KUMAR</title>
    <description>The latest articles on Forem by VISHAL KUMAR (@vk224).</description>
    <link>https://forem.com/vk224</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fthepracticaldev.s3.amazonaws.com%2Fi%2F99mvlsfu5tfj9m7ku25d.png</url>
      <title>Forem: VISHAL KUMAR</title>
      <link>https://forem.com/vk224</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://forem.com/feed/vk224"/>
    <language>en</language>
    <item>
      <title>AI Gatekeeper: A Highly Efficient Rejection-as-a-Service Platform 😂</title>
      <dc:creator>VISHAL KUMAR</dc:creator>
      <pubDate>Sat, 11 Apr 2026 18:42:10 +0000</pubDate>
      <link>https://forem.com/vk224/ai-gatekeeper-a-highly-efficient-rejection-as-a-service-platform-1kl0</link>
      <guid>https://forem.com/vk224/ai-gatekeeper-a-highly-efficient-rejection-as-a-service-platform-1kl0</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/aprilfools-2026"&gt;DEV April Fools Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What We Built&lt;/strong&gt;&lt;br&gt;
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.&lt;/p&gt;

&lt;p&gt;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.&lt;/p&gt;

&lt;p&gt;Demo&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Demo&lt;/strong&gt;&lt;br&gt;
Live app: &lt;a href="https://aigatekeeper.vercel.app/" rel="noopener noreferrer"&gt;https://aigatekeeper.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;

&lt;p&gt;The project is built as a tiny &lt;strong&gt;Vercel app&lt;/strong&gt; with exactly what it needs and nothing it deserves.&lt;br&gt;
Github : &lt;a href="https://github.com/Vk2245/DEV-April-Fools-Challenge" rel="noopener noreferrer"&gt;https://github.com/Vk2245/DEV-April-Fools-Challenge&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Frontend: pure HTML, CSS, and vanilla JavaScript in index.html&lt;br&gt;
Backend: Vercel serverless function in evaluate.js&lt;br&gt;
Routing: vercel.json&lt;br&gt;
Environment example: .env.example&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;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.&lt;/p&gt;

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

&lt;p&gt;Tech stack:&lt;/p&gt;

&lt;p&gt;Vanilla HTML, CSS, JavaScript&lt;br&gt;
Vercel Serverless Functions&lt;br&gt;
Gemini API&lt;br&gt;
Groq API fallback&lt;br&gt;
Local .env support for development&lt;/p&gt;

&lt;h2&gt;
  
  
  Prize Category
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Best Google AI Usage&lt;/strong&gt;&lt;br&gt;
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.&lt;/p&gt;

&lt;h2&gt;
  
  
  What Makes It Funny
&lt;/h2&gt;

&lt;p&gt;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.”&lt;/p&gt;

&lt;h2&gt;
  
  
  Team Members :
&lt;/h2&gt;

&lt;p&gt;Arnav Tiwari :&lt;a class="mentioned-user" href="https://dev.to/arnav_tiwari_91d5f9b0f961"&gt;@arnav_tiwari_91d5f9b0f961&lt;/a&gt; &lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>googleaichallenge</category>
      <category>ai</category>
      <category>machinelearning</category>
    </item>
  </channel>
</rss>
