AI Interview Question Bank

Pick a role. Answer out loud before you open the notes. Follow-ups and scoring are what the next interviewer uses.

AI / Agentic Systems Engineer

Ships agents that call tools, retrieve documents, and finish a job without looping forever. Interviewers want a loop you can debug, a retrieval path you can measure, and a cost number.

0/22 seen

Skills tested

Tool calling loopsRAG and hybrid searchLangGraph / custom runtimesEval sets and tracesSchema validationCachingTimeouts and human gates

Practice bank for AI hiring loops. Last updated August 2026.

AI Interview Question Bank

AI / Agentic Systems Engineer

What is ReAct, and when would you not use it?

Detail the sequence of system, user, assistant, and tool/function messages during an LLM tool-calling cycle.

When is a second agent worth the extra calls?

What is a vector database and why is it needed for RAG?

What is the purpose of a system prompt in an LLM agent, and how does it differ from a user prompt?

What is RAG (Retrieval-Augmented Generation), and when would you use it instead of only prompting the model?

What is an embedding, and how does it help search in a RAG system?

How do you evaluate a RAG system beyond reading a few sample answers?

A tool schema says location is a string. The model sends {"city":"Paris"}. What should your runtime do?

What is MCP, and when is it just another tool wrapper?

How do you stop an agent from emailing a customer without a human?

Plan-and-execute vs ReAct. When do you write the plan up front?