<?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>DEV Community: Doğukan DEMİR</title>
    <description>The latest articles on DEV Community by Doğukan DEMİR (@doukan_demir_e475427969a).</description>
    <link>https://dev.to/doukan_demir_e475427969a</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F2219162%2Fa55369a6-155b-45e0-bc57-7a9060525089.jpg</url>
      <title>DEV Community: Doğukan DEMİR</title>
      <link>https://dev.to/doukan_demir_e475427969a</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/doukan_demir_e475427969a"/>
    <language>en</language>
    <item>
      <title>How I built an AI interview prep tool with Next.js + Claude API (and what I learned)</title>
      <dc:creator>Doğukan DEMİR</dc:creator>
      <pubDate>Sun, 31 May 2026 11:08:15 +0000</pubDate>
      <link>https://dev.to/doukan_demir_e475427969a/how-i-built-an-ai-interview-prep-tool-with-nextjs-claude-api-and-what-i-learned-854</link>
      <guid>https://dev.to/doukan_demir_e475427969a/how-i-built-an-ai-interview-prep-tool-with-nextjs-claude-api-and-what-i-learned-854</guid>
      <description>&lt;p&gt;I've been job hunting for a while. Every interview meant &lt;br&gt;
2-3 hours of manual prep — researching the company, &lt;br&gt;
guessing questions, writing out answers.&lt;/p&gt;

&lt;p&gt;So I built Hiddea.&lt;/p&gt;

&lt;h2&gt;
  
  
  How it works
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;User pastes a job listing URL&lt;/li&gt;
&lt;li&gt;We scrape the listing + company page with Cheerio&lt;/li&gt;
&lt;li&gt;Claude API analyzes everything and generates:

&lt;ul&gt;
&lt;li&gt;Role-specific behavioral questions with STAR frameworks&lt;/li&gt;
&lt;li&gt;Technical questions based on the actual stack in the listing&lt;/li&gt;
&lt;li&gt;Company culture insights&lt;/li&gt;
&lt;li&gt;Personalized study plan&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Tech stack
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Next.js 14 (App Router) — web + API routes&lt;/li&gt;
&lt;li&gt;React Native (Expo) — mobile&lt;/li&gt;
&lt;li&gt;PostgreSQL + Redis — database &amp;amp; caching&lt;/li&gt;
&lt;li&gt;Claude API (Sonnet) — AI engine&lt;/li&gt;
&lt;li&gt;Drizzle ORM — type-safe queries&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  The interesting part — prompt engineering
&lt;/h2&gt;

&lt;p&gt;The hardest part wasn't the scraping or the UI. &lt;br&gt;
It was getting Claude to generate questions that are &lt;br&gt;
actually specific to the listing, not generic templates.&lt;/p&gt;

&lt;p&gt;Key insight: you have to explicitly tell the model to &lt;br&gt;
cross-reference the tech stack, seniority level, and &lt;br&gt;
company size — otherwise it defaults to generic output.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I learned
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;"Anyone can do this with ChatGPT" is the most common objection
→ True, but most people don't. The value is the setup, not the AI.&lt;/li&gt;
&lt;li&gt;Scraping job listings is messier than expected
→ LinkedIn blocks, Indeed has paywalls, Greenhouse is clean&lt;/li&gt;
&lt;li&gt;Prompt caching saves ~40% on Claude API costs&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What's next
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;CV upload + matching against job requirements&lt;/li&gt;
&lt;li&gt;Mock interview mode&lt;/li&gt;
&lt;li&gt;Multi-language support (already have 4 languages)&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Try it at hiddea.com — free, no credit card needed.&lt;br&gt;
Happy to answer questions about the implementation 👇&lt;/p&gt;

</description>
      <category>ai</category>
      <category>career</category>
      <category>nextjs</category>
      <category>showdev</category>
    </item>
  </channel>
</rss>
