Interview Prep
Practise the FDE interview, round by round
Real Forward Deployed Engineer and AI engineer rounds. Write your answer, get graded in seconds, see exactly what an interviewer would mark down.
15problems live, growing to the FDE 50
Free: 3 graded attempts a day. Unlimited with membership.
Five rounds, graded the way interviewers grade
Problems
- Use AI to clear a health-insurance claims backlogclarify before building · success metric · scope controlBreak it downEasy20m
- Cut 911 dispatch time with AIclarify before building · success metric · human in the loopBreak it downMedium30m
- Scope an AI copilot for 6,500 field techniciansclarify before building · success metric · stakeholder alignmentBreak it downHard40m
- Push back on scope creep in week 5 of an AI pilotscope control · stakeholder alignmentClient scenarioEasy15m
- Tell the CTO the AI pilot missed its targetbad news delivery · success metric · stakeholder alignmentClient scenarioMedium30m
- Unblock production data access at a bankdata access · PII constraint · stakeholder alignmentClient scenarioHard35m
- Design a VPC-only HR policy assistantprivate deployment · retrieval freshness · cost/latency budgetSystem designEasy25m
- Design PII-safe call summarisation for an Indian contact centrePII constraint · cost/latency budget · structured outputSystem designMedium35m
- Design an on-prem agent for pharma deviation investigationsprivate deployment · audit trail · human in the loopSystem designHard40m
- Route customer support tickets into five queuesclassification routing · prompt injectionPrompt & agent katasEasy15m
- Extract invoice fields as strict JSONstructured output · data validation · prompt injectionPrompt & agent katasMedium25m
- Document Q&A that resists prompt injection and masks PIIprompt injection · PII constraintPrompt & agent katasHard35m
- Debug an agent trace: unit mismatch in a freight quotetool-call failure · data validationDebug the traceEasy15m
- Debug a RAG trace: the assistant cited a superseded policyretrieval freshness · eval designDebug the traceMedium20m
- Debug an agent trace: a refund agent stuck in a looptool-call failure · agent loop controlDebug the traceMedium20m
A 4-week plan
Week 1
Warm up on every round
Week 2
Clients and design
Week 3
Katas and traces
Week 4
Hard rounds, timed
Practice shows on your profile. A credential proves it.
Problems you solve build a skill map on your public profile. Companies hiring through ADaSci see it next to your verified credentials.