Interview practice
AI Interview Question Banks
7 banks of AI engineering interview questions, each with plain-language answers and, where a number can be honestly computed, a worked answer showing the full arithmetic. Every bank links the glossary it leans on, and every answer names its assumptions.
Ten interview questions on serving AI models: KV cache sizing, batching, scheduling and throughput ceilings, each with a plain answer and a worked number.
Ten interview questions on 8-bit quantization and what tokens cost: memory math, precision trade-offs, break-even GPU decisions, each with a worked answer.
Ten interview questions on GPU selection for model serving: memory budgets, bandwidth ceilings and when a cheaper card wins, with worked numbers.
Ten interview questions on scaling AI serving: autoscaling signals, warm pools, spot capacity and cache-aware routing, with worked numbers.
Ten interview questions on AI response latency: time to first token, streaming, and the budgets users feel, with plain answers and worked numbers.
Ten interview questions on cache-aware request routing: round-robin hit rates, prefix copy arithmetic, hash concentration and locality, with worked arithmetic.
Ten interview questions on model API cost metering: per-request bills, cache discounts, retry re-billing and untagged spend, with answers and full arithmetic.
New to the vocabulary these questions lean on? The glossary defines each term first, with worked examples.