DEV Community

Doğukan DEMİR
Doğukan DEMİR

Posted on

How I built an AI interview prep tool with Next.js + Claude API (and what I learned)

I've been job hunting for a while. Every interview meant
2-3 hours of manual prep — researching the company,
guessing questions, writing out answers.

So I built Hiddea.

How it works

  1. User pastes a job listing URL
  2. We scrape the listing + company page with Cheerio
  3. Claude API analyzes everything and generates:
    • Role-specific behavioral questions with STAR frameworks
    • Technical questions based on the actual stack in the listing
    • Company culture insights
    • Personalized study plan

Tech stack

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

The interesting part — prompt engineering

The hardest part wasn't the scraping or the UI.
It was getting Claude to generate questions that are
actually specific to the listing, not generic templates.

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

What I learned

  • "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.
  • Scraping job listings is messier than expected → LinkedIn blocks, Indeed has paywalls, Greenhouse is clean
  • Prompt caching saves ~40% on Claude API costs

What's next

  • CV upload + matching against job requirements
  • Mock interview mode
  • Multi-language support (already have 4 languages)

Try it at hiddea.com — free, no credit card needed.
Happy to answer questions about the implementation 👇

Top comments (0)