Turn the rest of this series into a schedule: six weeks, about ten hours a week, covering progressive coding, LLM system design, ML fundamentals, the LLM practical round, a project presentation and the values round - plus what OpenAI, Anthropic and DeepMind say about using AI while you interview.
Every frontier lab has a round about why you want to work on AI and how you think about its risks, and it has a real bar. What OpenAI, Anthropic and DeepMind ask, the four answers that sink candidates, and a structure for answers that are specific, honest and yours.
Some AI-lab loops now include a round where you work with a model through an API in a notebook. It is not a prompt-writing contest. It tests whether you can call a model a thousand times safely, cheaply and measurably. The patterns that pass, with code.
Design a chat-completion service and the scarce resource is not CPU or database connections - it is accelerator memory and bandwidth. KV-cache sizing, prefill versus decode, continuous batching, TTFT and TPOT, and token-based admission control, worked with numbers.
Scaled dot-product attention fits in ten lines of NumPy. The interview is the follow-ups: why divide by the square root of d, how to make softmax stable, where the causal mask goes, what the KV cache saves, and why grouped-query attention exists. A complete walkthrough.
One problem, four levels, ninety minutes, and the requirements change every twenty. A full worked solution to the in-memory database problem family that candidates report from AI-lab assessments - and the three design choices in level one that decide whether you finish.
DeepMind runs closer to Google than to a startup lab: algorithm-style coding where the code must run, a maths-and-ML fundamentals round candidates call the quiz, ML design as the hardest part of the loop, and a hiring committee at the end. What each round tests and how to prepare.
A four-level coding assessment where level one decides whether you finish level four, a hiring-manager deep dive, an onsite with an LLM practical round, a values round with a real bar - and an explicit policy on when you may use Claude. What candidates report and how to prepare.
What OpenAI's own interview guide says, what candidates report, and how to prepare for each stage - the skills assessment, the four-to-six-hour final loop, the senior project presentation, and the rule on AI tools that changes from round to round.
The stages look familiar - recruiter call, assessment, a loop of four to six rounds. What happens inside them does not. Practical coding instead of puzzles, system design with a GPU-bound model in the middle, and a mission round that is scored, not decorative. Here is where the prep time should go.