Implementation problems where you write real AI engineering code. Tests production coding skills specific to LLMs, RAG, and agents. Seen at: technical onsite rounds at Anthropic, OpenAI, Meta.
Conceptual questions asked in screening rounds. Tests whether you understand AI fundamentals and can explain them clearly. Seen at: most AI companies in phone screens.
Architecture questions asking you to design a full AI system under pressure. Tests whether you can think in systems, not just features. Seen at: senior and staff engineer onsites at all major AI labs.
Debug broken code using an AI assistant. Google and Meta's new interview format — tests prompt quality, catching AI mistakes, and debugging skill. Seen at: Google (2026), Meta (2025).