Interview practice by role
AI Engineer Interview Practice
AI engineer interviews centre on building reliable applications on top of language models. Retrieval quality, evaluation design and cost control come up more often than model theory.
160 practice questions4 topics5 typical rounds
How the interview usually runs
Loops differ by company. Treat this as a common pattern, not a fixed format.
- 01Screening callWhat you have built with LLMs, which models you used and what went wrong.
- 02Python and LLM fundamentalsPrompting, structured output, tokens, context limits and calling model APIs robustly.
- 03RAG and evaluation deep diveChunking, embeddings, retrieval quality, and how you measure whether answers are correct.
- 04Applied design roundDesign an assistant or agent: tools, memory, latency, cost and safety.
- 05Behavioural roundShipping under uncertainty and explaining model limits to non-technical people.
What interviewers probe
- How you would evaluate an LLM feature before and after release
- Retrieval design: chunking, ranking and when retrieval fails
- Prompt injection and other safety risks in tool-using systems
- Cost and latency trade-offs between models
- When not to use an LLM at all
Common mistakes
- Describing a demo without saying how you measured quality
- Treating the model as deterministic and skipping retries and fallbacks
- Naming a framework instead of explaining the underlying design
Practice questions for AI Engineer interviews
A starting set from each topic. Every question has a short answer, the full reasoning and the follow-ups interviewers ask.
01How do temperature, top-p and sampling affect output?Mediumsampling · temperature · determinism02How do you get reliable structured output from an LLM?Mediumstructured output · json · validation03How do you manage context in a long-running application?Mediumcontext · memory · summarisation04RAG vs fine-tuning: how do you choose?Mediumrag · fine-tuning · trade-offs05When is RAG the wrong approach?Mediumrag · design · trade-offs06RAG vs long context windows: what actually differs?Mediumrag · long context · cost07Offline vs online evaluation?Mediumevals · online · offline08What makes a good golden dataset?Mediumdataset · evals · labels09How do you pick metrics for a specific task?Mediummetrics · evals · design10Why does mutability matter in Python?Mediummutability · aliasing · defaults11How do generators work and why use them?Mediumgenerators · laziness · yield12What is the iterator protocol?Mediumiterators · protocols · for loops
Reading is step one
Now say it out loud.
Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live AI Engineer mock interview and get scored feedback on depth, clarity and confidence.