All styles
Playbook · Interview style
AI engineering
The screen AI-native teams actually run: not model math, but the engineering around LLMs — structured output, retrieval, evaluation, and the failure modes that come with non-determinism. It's about shipping and operating AI features, not training models.
- LLMs
- RAG
- Evaluation
- APIs
Interview flow
A style plays out as a sequence of sessions — start each one on its own; your readiness rolls up across them.
- 1. Depth interviewAn AI-engineering depth round: prompting, RAG, evals, and safety.Start below
- 2. Code challengeHarden a naive LLM call — validation, retries, injection.
What interviewers usually probe
- Getting structured output reliably, and what you do when it's malformed.
- RAG quality — retrieval, grounding, and honest refusal.
- How you'd evaluate an LLM feature and catch a regression.
- Prompt injection and the untrusted-input boundary.
What we assess
- LLM application patterns and structured output
- Retrieval-augmented generation and grounding
- Evaluation and regression-catching for AI features
- Safety, cost, and latency in production
What we don’t
- Model training / research MLThis style is applied AI engineering. For modelling fundamentals, try the ML engineering focus session.
- Algorithmic (DSA) roundsNo coding puzzles here.
Practise this style
Run an adaptive ai engineering interview — pick your level, answer by voice or text, and get corrected like a senior engineer. Your first interview is free.
Practise AI engineering