AI engineer interview questions
An AI engineer interview is not a machine learning interview with new vocabulary. Nobody asks you to derive backpropagation. What gets tested is whether you can build on models that are confidently wrong and prove the result works: scoping a vague problem, choosing retrieval over a bigger prompt, deciding when an agent is the wrong answer, and saying how you would measure any of it. 853+ questions across 10 topics, each ordered easy to hard.
What the rounds actually test
- ▸The design round. Says architecture. Tests whether you name the constraint before drawing boxes, and whether you can pick the boring option and defend it.
- ▸The coding round. Says write code. Tests production habits: messy input, failure paths, output a program can rely on. Rarely an algorithm puzzle.
- ▸The evaluation question. Says metrics. Tests whether you have operated a system. 'How would you know it works' ends more of these interviews than any hard algorithm.
- ▸The behavioural round. Says tell stories. Tests judgment and restraint. The rarest good answer is the time you argued against using a model.
Practise by topic
LLM Interview Questions104Tokens, attention, context windows, sampling, prompting vs RAG vs fine-tuning, hallucinations, evals, cost and latency: the conceptual core every Applie...RAG and AI Agent Interview Questions92Retrieval pipelines, chunking, hybrid search, reranking, tool-using agents, guardrails, multi-tenancy and eval harnesses: the modal Applied AI design ro...AI Engineer Coding Interview Questions131Practical builds (parsers, in-memory stores, rate limiters, streaming) plus the LeetCode-medium staples, calibrated to the practical coding screens appl...Machine Learning Interview Questions121Classical ML, embeddings, metrics (precision/recall, AUC, calibration), experiment design and statistics: the modeling fundamentals applied AI roles sti...SQL and Data Engineering Interview Questions58Window functions, gaps-and-islands, Spark internals, lakehouse design, idempotent pipelines and CDC: the data plumbing under every production AI system.AI System Design Interview Questions109Productionizing a POC: latency and throughput budgets, idempotency and retries, caching, observability for non-deterministic systems, and private or air...MLOps Interview Questions55CI/CD for models, drift detection and retraining, model registries, Kubernetes inference, feature stores, and staging-to-production promotion: what ever...ML Infrastructure and GPU Interview Questions65GPU/TPU memory, distributed training and parallelism, quantization, inference serving (vLLM, batching, KV cache), and scaling API gateways: the infra de...AI Security Interview Questions61Prompt-injection defense in depth, PII handling, tenant isolation, abuse and consumption controls, audit trails, and SOC 2 / EU AI Act questions enterpr...Behavioural Interview Questions for AI Engineers57Owning ambiguous ML projects end to end, the model-failure post-mortem, cross-functional trade-offs, and the deep-dive on a system you actually shipped:...
Not sure where to begin? Follow the map: four stages of preparation, and where to join them given what you already know. If agents are your weak spot specifically, the agent questions are collected on their own page.