AI Engineering Interview Questions
Practice with real interview questions. Each includes a think-first prompt, a strong answer, weak-answer indicators and follow-up questions.
What is RAG and how does it work? Walk me through the full pipeline.
When would you use RAG vs fine-tuning? Can you use both?
What is an AI agent? How is it different from a regular LLM call?
Your agent keeps calling the same tool with paraphrased arguments and never converges. How do you fix this?
How do you evaluate a RAG system? What metrics matter?
What is LLM-as-judge? What biases does it have and how do you mitigate them?
Your AI feature costs $4,800/day. How do you cut costs by 80% without hurting quality?
How do you defend a RAG system against prompt injection via retrieved documents?
Design a RAG system for a bank with 60,000 internal documents and 18,000 employees with role-based access control.
What does a Forward Deployed Engineer do? How is it different from a regular software engineer?
A customer says 'we want an AI assistant over our docs.' What's your first 30 minutes with them?
What are embeddings? How do they work and why are they useful?
What is function calling (tool calling)? How does it work under the hood?
A hospital network says: "Use AI to reduce emergency department wait times." Walk me through how you would approach it.
An insurer has a 40,000-claim backlog and wants AI to clear it. How do you break the problem down?
A retailer says returns fraud is costing them millions. They ask what AI can do. Where do you start?
A manufacturer wants to "use GenAI to reduce machine downtime". What questions do you ask and what do you propose?
A city council takes 90 days to approve building permits and wants AI to speed it up. How do you approach it?
A bank wants to deploy an always-on AI agent for its procurement team. Design the rollout.
In your first meeting, the VP says: "Just show us a demo next week and we will decide." How do you respond?
The customer says the system must be 100% accurate or it is useless. How do you handle that?
Three weeks in, the customer's security team says the project cannot send any data to an external model. What do you do?
Your pilot is technically working but users are not using it. What do you do?
Midway through a POC, the customer keeps adding "small" requests. How do you manage it?
An executive asks: "Why does the AI make things up, and how do I know it won't do it here?" Answer as you would to them.
Tell me how you would tell a customer that their favourite AI use case should not be built.
Design a document assistant for 20,000 employees where people must only see answers from documents they are allowed to access.
A defence customer requires everything on-premises with no internet. How does your architecture change?
Your POC impressed everyone. List what has to change before it can go to production.
Design the approval policy for an AI agent that can read and send email on behalf of a sales team.
The customer's CFO asks what the AI system will cost per month at full rollout. How do you estimate it?
Users report the assistant's answers got worse this week. Nothing in your code changed. How do you investigate?
How do you build an evaluation set with a customer who has never done it?
You join a customer's project and must add a feature to a large codebase you have never seen, by tomorrow. How do you work?
The key system has no API, only a 15-year-old web interface. How do you integrate an AI workflow with it?
Your agent reads customer emails. How do you protect against prompt injection hidden in those emails?
Tell me about a project that failed. What did you learn?
Describe a time you had to deliver with very little direction.
Why do you want to be a Forward Deployed Engineer rather than a product engineer?
Tell me about a time you disagreed with a customer or stakeholder. How did it end?
Explain RAG to a non-technical operations manager in under a minute.