Inference Engineering Interview Prep Pack with Roni Das

120 interview questions with full answers and a diagram for every question. One PDF: 292 pages and 141 diagrams.

What it covers

Serving LLMs: GPU memory, the KV cache, static and continuous batching, paged attention, quantization, speculative decoding, tensor, pipeline and expert parallelism, serving stacks, latency versus throughput, and cost per token.

What every question includes

The question, as it is asked in real interviews, with its level: junior, mid, senior or staff.

What the interviewer is really testing.

A full model answer in plain English. The first sentence answers the question; the detail follows.

A diagram that shows the core idea of the answer.

The key points the interviewer listens for.

Likely follow-up questions, each with a hint for the answer.

Common mistakes and traps.

Also inside

A glossary of the terms used in the pack.

Questions ordered from small ideas to big ones: 24 junior, 42 mid, 36 senior and 18 staff.

Format

One PDF, delivered here right after purchase. The questions are grouped into sections that go from small ideas to big ones, so you can study in order or jump to a topic. Written by Roni Das in original words, in simple English for readers whose first language is not English.