Active recall / Make the idea stick
Close the book.Open your memory. Explain a concept or close a design in 60 seconds. Reveal the notes, compare your answer, and choose what to revisit.
TopicMixed — all topics AI System Design Interview Practice AI Engineering and System Design Question Bank Answer Frameworks for AI System Design Interviews Common Pitfalls in AI System Design Interviews Behavioral Interviews for AI Engineers and Engineering Leaders AI Roles, Hiring Evidence and Interview Preparation — September 2026 AI Engineering FAQ LLM Internals: How a Model Learns and Answers Tokenization Deep Dive: How Text Becomes Model Input Attention Mechanisms: How Tokens Share Information Transformer Architecture: From Recurrent Memory to Modern AI Embeddings and Vector Spaces Inference Pipeline Model Taxonomy Capability Assessment Pricing and Costs Model Selection Guide Pretraining: learning a reusable language model Fine-tuning: change behavior for a measured reason LoRA and QLoRA: understand what becomes smaller Preference learning: RLHF and DPO Knowledge distillation: transfer useful behavior, measure the loss Synthetic data: create examples for a specific learning gap Quantization: budget memory without guessing quality RLVR and GRPO: train against a checkable outcome Inference: follow the request before optimizing it KV and prefix caches: reuse computation with clear boundaries Speculative decoding: propose cheaply, verify correctly Batching: schedule useful work without hiding the wait PagedAttention: allocate the cache as the sequence grows Serving infrastructure: build a service around the model AI cost optimization: improve the economics of a completed task Diffusion language models: parallel refinement with a measurable contract Local and edge inference: design for the device and the trust boundary Prompt Engineering Fundamentals Few-Shot and In-Context Learning Chain-of-Thought Prompting and Reasoning Tree of Thoughts and Deliberate Search Context Engineering Structured Generation Prompt Optimization with DSPy Prompt Injection and Defense RAG Fundamentals Chunking Strategies Embedding Models Vector Databases and Search Indexes Hybrid Search Reranking Strategies GraphRAG Agentic RAG Advanced Retrieval Patterns Contextual Retrieval Late Interaction and ColBERT Multimodal RAG RAG Evaluation Patterns Production RAG at Scale Data Engineering for AI Agentic Systems: Design, Control and Verification Agent Fundamentals Reasoning Loops: ReAct and Beyond Tool Use and MCP Multi-Agent Orchestration Agent Memory and State Planning and Decomposition Error Handling and Recovery Human-in-the-Loop Patterns Agentic Security and Sandboxing Evaluating Agentic Systems Durable Execution for Long-Running Agents Loop Engineering Memory Architectures Short-Term Context Management Long-Term Memory Agentic Memory with Mem0 Semantic Caching State Management Patterns LangChain Deep Dive LangGraph Orchestration LangSmith Observability LlamaIndex: document retrieval and event-driven workflows DSPy: programming and evaluating model behavior Semantic Kernel and Microsoft Agent Framework Multi-agent frameworks: CrewAI, AutoGen, and current SDKs Choosing an AI framework: requirements, evidence, and operating cost Claude Code: designing a dependable coding workflow Coding models and agents: choose by evidence Pydantic AI and Mastra: typed boundaries for agent applications Framework changes: reproduce, diagnose, migrate OCR and layout analysis: turn documents into reliable evidence LLM infrastructure: size work, protect deadlines, recover failures CI/CD for LLM applications: release the complete behavior AI gateways and model routing: enforce policy before choosing a model FinOps and token economics: measure cost per useful outcome LLM security: protect data, authority and execution Access control for AI applications Guardrails: enforce a specific rule at the right boundary Ensembles: use multiple outputs only when they improve the decision Reliability patterns: bound work and recover without duplicate effects AI governance and compliance: turn obligations into operated controls LLM evaluation: measure the behavior required by the product AI observability: explain what happened to a user task Benchmarks and leaderboards: compare evidence before choosing a model AI Design Patterns: Choose the Mechanism That Solves the Failure AI Anti-Patterns: Diagnose the Assumption Before Replacing the Design Design an Enterprise Knowledge Assistant Design a Conversational Customer-Support Agent Design a Source-Verified Financial Research Assistant Design an IDE Code Assistant Design a Content-Moderation Platform Design a Fresh Market-Intelligence Search Service Design an Agent That Produces a Reviewable Code Change Design a Multi-Tenant Contract-Analysis Platform Design Support Automation That Resolves the Right Issue Design a Contract Document-Intelligence Pipeline Design Movie Recommendations with Truthful Explanations Case Study: Pharmaceutical Promotion Review Case Study: Clinical Voice Documentation Case Study: Real-Time Payment Fraud Decisions Case Study: Enterprise Knowledge Assistant Case Study: Expense Operations with a Computer-Use Agent Case Study: Customer-Specific Fine-Tuning Platform Design an Evaluation Gate for AI Releases Design a Customer-Specific Distillation Pipeline Design an Enterprise Knowledge Agent with MCP Tool-use agents: choose the execution model before the product Architecture patterns for dependable tool-use agents OpenClaw: designing a persistent assistant around a trusted gateway Computer-use agents: from screen observations to verified outcomes Building tool-use agents: contracts, execution, and evidence Tool-agent use cases: choose the workflow, prove the value Safety and governance for tool-using agents Real-time voice agents: conversation, timing, and trustworthy actions Multimodal Generation: From Prompt to Publishable Media AI Architecture Pattern Reference AI Evaluation Lab: Build, Inspect, and Compare AI Evaluation: Evidence, Metrics, and Release Decisions Research Reading for AI System Design Design an Adaptive AI Learning Tutor Design an AI Gateway and Model-Routing Service Design a Deadline-Aware Batch Inference Platform Design an Image and Video Generation Platform Design a Multi-Tenant Model-Serving Platform Design a Multi-Tenant Vector Search Service ShowAll cards Due for review New cards
← Previous Reveal notes Next →
Review intervals: Again → now; Good → 1, 3, 7, 14, then 30 days; Easy advances two intervals, up to 30 days. This is a simple study schedule, not a validated memory assessment.
