AI System Design
An end-to-end AI system on a whiteboard: retrieval, agents, evaluation, scale, security and cost.
- Format
- 45 to 60 minutes, whiteboard
- Study
- 4 modules
- Practise
- 42 lessons
- Mocks
- 6
- By
- Sourav Banerjee
What It Tests
A strong answer shows all three.
- 1You pin down users, data and the success measure before drawing boxes.
- 2You name the trade-offs in retrieval, model choice, latency and cost, with numbers.
- 3You show how the system is evaluated, secured and operated after launch.
Study in Learn
The modules that teach what this round tests. Start here if a topic is new.
Practise
Answer each question out loud before you read the model answer.
Design questions (6)
- Practise Enterprise RAG Interview QuestionsEnterprise RAG in production
- Practise Security and Tenancy Interview QuestionsSecurity, tenancy and prompt injection
- Practise Observability and Diagnosis Interview QuestionsObservability and production diagnosis
- Practise Cost and Latency Interview QuestionsCost and latency optimisation
- Practise Scaling and Reliability Interview QuestionsScaling and reliability
- Practise AgentOps Interview QuestionsAgentOps: versioning, rollout, CI/CD, red teaming
Depth questions (12)
- Why RAG & the RAG Lifecycle: interview questionsWhy RAG & the RAG Lifecycle
- Chunking Strategies: interview questionsChunking Strategies
- Retrievers & Hybrid Search: interview questionsRetrievers & Hybrid Search
- LangGraph Core: Graphs, State & Control Flow: interview questionsLangGraph Core: Graphs, State & Control Flow
- The Agent Loop & Tool Calling: interview questionsThe Agent Loop & Tool Calling
- Practise Tool Design Interview QuestionsTool design and the agent-computer interface
- Practise Context and Orchestration Interview QuestionsContext engineering and orchestration
- Long-Term & Agentic Memory: interview questionsLong-Term & Agentic Memory
- Practise Robust Agent Interview QuestionsRobust agents: retries, fallbacks and guards
- Multi-Agent Architectures: interview questionsMulti-Agent Architectures
- Practise Agentic Workflow Platform Interview QuestionsAgentic workflow platforms
- Practise Evaluation and Release-Gate Interview QuestionsEvaluation and release gates
Staff platform questions (12)
- Build the MVP: think like a staff engineerBuild the MVP
- Distributed execution: think like a staff engineerDistributed execution
- Enterprise knowledge: think like a staff engineerEnterprise knowledge
- Scale the application: think like a staff engineerScale the application
- Handle overload and fair scheduling: think like a staff engineerHandle overload and fair scheduling
- Enterprise security: think like a staff engineerEnterprise security
- Reliability and consistency: think like a staff engineerReliability and consistency
- Production AI quality and releases: think like a staff engineerProduction AI quality and releases
- Inference and GPU infrastructure: think like a staff engineerInference and GPU infrastructure
- Fine-tuning and model customisation: think like a staff engineerFine-tuning and model customisation
- Global multi-region platform: think like a staff engineerGlobal multi-region platform
- Staff-level platform architecture: staff interview practiceStaff-level platform architecture
System design worksheets (12)
- Enterprise Knowledge Assistant with RAGSystem Design Practice
- Secure Multi-Tenant AI PlatformSystem Design Practice
- Natural-Language-to-SQL Analytics AssistantSystem Design Practice
- Customer Support AutomationSystem Design Practice
- AI System for an Air-Gapped EnvironmentSystem Design Practice
- Evaluation and Release GatingSystem Design Practice
- High-Volume Batch Inference PipelineSystem Design Practice
- Tool-Using Agent with Safety ControlsSystem Design Practice
- Reliable Workflow Orchestration SystemSystem Design Practice
- Observability for a Customer-Facing AI ApplicationSystem Design Practice
- Configurable Platform for Customer-Specific WorkflowsSystem Design Practice
- Enterprise Chatbot PlatformSystem Design Practice
Mock Interview
Run it under time with a partner, then score it against the scorecard.
- Running and Scoring a Mock InterviewMock interviews and scorecards
- Internal Knowledge Assistant Mock InterviewMock interviews and scorecards
- Agentic Support Workflow Mock InterviewMock interviews and scorecards
- Financial Compliance and Red-Team Mock InterviewMock interviews and scorecards
- SRE Triage Agent Mock InterviewMock interviews and scorecards
- Multi-Tenant SaaS Assistant Mock InterviewMock interviews and scorecards