Start free
AI infra interview prep for software engineers

Crack the AI infra system design interview.

Watch a full mock interview of Design ChatGPT, a problem candidates report from OpenAI. Answer each hard question yourself, then see the staff-level answer.

Watch a free interview → no ML background needed: your backend skills carry over ✓
Design ChatGPT · 11:26
MeiArjun
Mei, the interviewer, asks:

A worker freezes. Another takes over. Then the first one wakes up. What now?

Most prep gives you a write-up.
We give you the interview.

watch

the full interview

questions, wrong turns, recovery
answer

at every pause

the video stops at each hard question
compare

mid · senior · staff

see the bar for your level

What carries over, and what's new.

AI infra is distributed systems with a model in the middle

what you knowwhat it becomes in AI infra
Load balancingrouting requests to GPU replicas by queue depth and KV-cache fit
Queues and backpressurecontinuous batching and admission control on GPUs
Cachingprompt caching and the KV cache in GPU memory
Rate limitingtoken budgets per user and per tenant
Streaming over websocketsstreaming tokens, and resuming a stream that dropped
Search indexesvector retrieval (RAG) with citations
Leases and idempotent writesfencing a GPU worker that froze and woke up
Monitoring and A/B testsevals that gate every model change

what's actually new: tokens · KV cache · GPU memory · batching · retrieval · evals. The interviews teach them as they come up.

the problems AI loops ask

    See the bar for your level.

    The finished whiteboard from Design ChatGPTyours to edit ↗
    available today · free

    Design ChatGPT

      coming next to the library +

        Costs less than one hour with a coach.

        Pro includes every interview in the library and each new one we add · every pause with the mid, senior and staff answers · breakdowns and searchable transcripts

        expense it: email your manager → or start free: one full interview →

        Questions

        Who is it for?

        Software engineers (backend, infra, platform, full-stack) going after AI engineer and AI infrastructure roles, from mid to staff. Each interview is broken down for every level.

        Do I need an ML background?

        No. AI infra rounds are distributed systems with a model in the middle, so most of what you know carries over: load balancing, queues, caching, rate limits, streaming, idempotency. What's new is a short list: tokens and context windows, the KV cache and GPU memory, batching, retrieval and evals. The interviews teach those as they come up, on top of what you already know.

        Who are Mei and Arjun?

        Our interviewer and candidate. Every interview is written by us from candidate reports, reviewed, and produced with AI voices, which is how we ship new problems and every level of each.

        Does it renew automatically?

        Monthly renews until you cancel. Yearly renews once a year, and we email you 7 days before it does. If you subscribed at an intro price, you keep it for as long as you stay subscribed. Lifetime is paid once. Any plan: full refund within 30 days of your first payment, no questions asked; just reply to your receipt.

        How does the intro offer work?

        New members get Pro at $79/year or lifetime at $199 for a few days after their first visit; the bar at the top shows exactly how long you have. When it ends, it's $119/year and $299 lifetime for you, and it doesn't come back. Get Pro before then and you keep your price for as long as you stay subscribed. Monthly is always $39.

        Can my company pay?

        If your employer reimburses professional development, use "expense it" under the plans: it opens a short note to your manager with the price and the refund terms.

        See where your answer lands