Forward Deployed Engineer interviews test more than coding. The round that catches good engineers out is the customer scenario: you are handed a vague business problem and have to work out what to build. This page covers the usual rounds, a way to answer that one, and 15 scenarios to practise.

The usual rounds

Every company runs its own process, so check the interview guide for the one you are applying to. Most FDE loops include these 4 kinds of round:

1. Coding. Practical problems in Python or TypeScript: parsing data, calling APIs, fixing a bug. Less puzzle-style than a typical software interview.
2. System or integration design. Design how an AI system connects to a customer’s existing data and tools.
3. Customer scenario, sometimes called the decomposition round. A vague problem, and you break it down.
4. Behavioural. How you handled unclear requirements, a difficult stakeholder or a project that went wrong.

How to answer a customer scenario

The mistake is jumping straight to a solution. Use these 5 steps, out loud:

1. Ask before you answer. Who has this problem? How do they handle it today? What does it cost them in time or money?
2. Define success. Agree on one number that shows it worked, such as “replies in under 2 minutes”.
3. Find the data. Where does the information live, who owns it, how messy is it, and are you allowed to use it?
4. Propose the smallest first version. Something you could put in front of the customer in 2 weeks, with a human checking the AI’s output.
5. Name the risks. What could go wrong, how you would catch it, and what you would build next.

15 practice scenarios

I wrote these for practice. They are not taken from any company’s real interviews. Take one, set a 30-minute timer, and talk through the 5 steps.

1. A hospital chain says its doctors spend too long writing discharge summaries.
2. A bank’s support team takes 2 days to answer customer emails.
3. An e-commerce company wants to “use AI to reduce returns”.
4. A law firm wants lawyers to find clauses across 50,000 old contracts.
5. A logistics company’s dispatchers plan truck routes by hand in spreadsheets.
6. An insurance company wants to speed up claim approvals without approving more fraud.
7. A coaching institute wants an AI tutor that answers student doubts at night.
8. A factory wants to predict which machines will break down next.
9. A sales team says its CRM data is too messy for any AI tool to use.
10. A customer’s pilot chatbot gives wrong answers 1 time in 10, and they want to cancel.
11. A hotel group wants one AI agent to handle bookings on WhatsApp, email and phone.
12. A government office wants citizens to ask questions about schemes in Hindi and get correct answers.
13. An HR team gets 5,000 resumes for each opening and wants help shortlisting fairly.
14. The customer’s IT team refuses to let any data leave their own servers.
15. Two leaders at the customer want different things from the same project, and you have 6 weeks.

Before the interview

• Prepare 3 stories from your own work: an unclear problem you made clear, a customer or stakeholder you won over, and a failure you fixed.
• Practise scenarios with a friend who plays the customer and gives short, vague answers.
• Read the company’s product docs, so your designs use what they actually sell.

New to the role? Start with the FDE Roadmap. This page is general career information, not a promise of a job.