How it works

From “my interview is in six weeks” to walking in ready

crackAI.dev is not a pile of resources. It is one loop: tell us the role and the date, get a plan built around it, learn and practice against that plan, then test yourself in a real mock interview — and let the debrief reshape what you study next.

  1. Step 1 · Start here

    Tell us the role and the deadline

    A two-minute onboarding: the role you are targeting, your experience level, and when the interview is. Add the real interview — company, date, job description — and everything downstream is grounded in that.

    • Role + experience level
    • Target date, or a real interview with its JD
    • Optional resume upload for grounded questions
    Set up your prep
  2. Step 2 · Your plan

    Get a dated study path

    We turn that into a day-by-day plan that fits the time you actually have. Topics, questions, labs, playgrounds, and mock checkpoints land on a calendar, ordered so the fundamentals come before the things built on them.

    • A real calendar, not a checklist
    • Core / important / optional tiers so you know what to cut
    • Reschedules itself as you complete work
    See your path
  3. Step 3 · Understand it

    Learn the theory

    24 topics covering what these interviews are actually built on — transformers, RAG, distributed training, RLHF, serving, quantization, agents — with tradeoffs, common traps, and revision notes. Highlight any passage and ask AI about it.

    • 24 topics + 50 foundational papers
    • Highlight-to-ask-AI on any passage
    • Cheat sheets and common traps per topic
    Browse topics
  4. Step 4 · Build intuition

    Play with the concepts

    Reading about attention is not the same as watching softmax collapse when you shrink the temperature. Twelve interactive explorers let you move the dials and see the tradeoff move — the fastest way to make a concept stick.

    • Self-attention, RoPE, KV cache, MoE routing
    • PPO clipping, GRPO advantages, bandits
    • Every number is live and reproducible
    Open the playground
  5. Step 5 · Do the work

    Practice on real problems

    Design the architectures these roles depend on, then build them. System design labs run on an interactive canvas with multi-level progression and automated scoring; coding labs run in a real Python sandbox with auto-graded tests and an AI assistant beside you.

    • System design labs with scored reference solutions
    • Coding labs in a real Python sandbox
    • Curated interview questions with company tags
    Start practicing
  6. Step 6 · Prove it

    Sit a real mock interview

    A fully spoken interview with an AI interviewer that listens, follows up on what you actually said, and grounds its questions in your resume and the job description. At the end you get a written debrief: what landed, what did not, and what to fix.

    • Natural turn-taking — it lets you think
    • Follow-ups grounded in your resume and JD
    • A detailed written debrief afterwards
    Try a mock interview

Then the loop closes. Your debrief shows what to fix, your path reshuffles around it, and you go again — with the weak spots moved to the front.

Start free

Starter includes 1 full voice mock interview a month with its debrief, free resume reviews, every topic, paper, and playground, and unlimited Design Lab scoring. No card required.

Build my pathCompare plans