Research Collections
⌘K
Phase 0 · The AI Engineer on the Edge1 lesson
What an AI engineer ships — and why the Cloudflare Workers runtime is the substrate for the whole roadmap
02Start hereThe AI Engineer's Roadmap: Skills, Tools & Career Path (2025+)Mid30m→Phase 1 · Models & Inference on Workers AI6 lessons
Pre-trained models and how inference actually runs — served from the edge via the Workers AI binding
04Start hereTransformer Architecture: Attention, Positional Encoding & ScaleMid20m→05Tokenization: BPE, SentencePiece & Vocabulary DesignMid20m→06LLM Architectures Compared: GPT, Claude, Llama, Gemini, Mistral, and BeyondMid24m→07Scaling Laws: Chinchilla, Emergent Abilities & Compute-Optimal TrainingMid20m→08Inference Optimization: KV Cache, Quantization & Speculative DecodingMid24m→09Pre-training: Data Curation, Objectives & CurriculumMid23m→Phase 2 · Prompting & Structured Output9 lessons
Talk to Workers AI models reliably — prompts, tool/function calling, and JSON-shaped output
10Start herePrompt Engineering Fundamentals: Principles & PatternsMid22m→11Few-Shot & Chain-of-Thought PromptingMid22m→12System Prompt Design: Instructions, Personas & GuardrailsMid25m→13Structured Output: JSON Mode, Tool Schemas & Constrained DecodingMid25m→14Function Calling & Tool Integration: APIs, Schemas & ExecutionMid21m→15Tool Use: Function Calling as an Agentic PatternMid4m→16Prompt Optimization: DSPy, Automatic Prompt Engineering & Meta-PromptingMid26m→17Prompt Caching: KV Cache Mechanics, Prefix Caching & Cost OptimizationMid41m→18Adversarial Prompting: Jailbreaks, Injections & Defense StrategiesMid25m→Phase 3 · Embeddings & RAG on Vectorize8 lessons
Connect models to your data with langchain-cloudflare embeddings and the Vectorize vector store
20Start hereEmbeddingsMid12m→21Embedding Models: Representations, Similarity & Fine-tuningMid21m→22Vector Databases: Indexing, ANN Search & Production PatternsMid25m→23Chunking Strategies: Splitting, Overlap & Semantic BoundariesMid23m→24Retrieval Strategies: Hybrid Search, Reranking & HyDEMid24m→25RAG: Retrieval-Augmented Generation as an Agentic PatternMid4m→26Advanced RAG: Agentic, Graph-Based & Multi-Hop RetrievalMid26m→27RAG Evaluation: Faithfulness, Relevance & Failure ModesMid29m→Phase 4 · Agents, Memory & Orchestration15 lessons
Agents that reason and act — LangGraph checkpointed on D1, long-term memory on Vectorize, durable on Workers
30Start hereAgent Architectures: ReAct, Plan-and-Execute & Cognitive FrameworksMid24m→31Multi-Agent Systems: Orchestration, Delegation & CommunicationMid21m→32Agent Memory: Short-term, Long-term & Episodic Memory SystemsMid24m→33Agent Orchestration: Routing, Handoffs & Supervisor PatternsMid38m→34Agent Harnesses: Event Loops, Permission Models & Tool SandboxingMid44m→35Agent SDKs: Claude, OpenAI, Vercel AI SDK & Framework ComparisonMid40m→36Code Generation Agents: Sandboxing, Iteration & Self-RepairMid24m→37Agent Debugging & Observability: Tracing, Replay & Root Cause AnalysisMid42m→38LangGraph: Stateful Multi-Agent Graphs for Production AIMid18m→39Memory: Persistence as an Agentic PatternMid5m→40Context Engineering: Designing What LLMs SeeMid16m→41Context Window Management: Token Budgets, Prioritization & Overflow StrategiesMid37m→42Memory Architectures for LLM Systems: Working, Episodic & Semantic MemoryMid38m→43Dynamic Context Assembly: Runtime Composition for LLM ApplicationsMid37m→44Context Compression: Fitting More Signal into Fewer TokensMid38m→Phase 5 · Evals, Safety & Observability18 lessons
Measure what matters and ship safely — evals, red-teaming, guardrails, and AI Gateway observability
45Start hereLLM Evaluation Fundamentals: Metrics, Datasets & MethodologyMid29m→46Benchmark Design: Contamination, Saturation & Domain-Specific EvalsMid21m→47LLM-as-Judge: Automated Evaluation, Calibration & BiasMid35m→48Human Evaluation: Annotation Design, Inter-Rater Reliability & ScaleMid21m→49Eval Frameworks Comparison: DeepEval, Promptfoo, RAGAS, Braintrust, LangSmith & MoreMid11m→50DeepEval Synthesizer: Synthetic Golden Generation for LLM EvaluationMid20m→51Agent Evaluation: Reliability, Tool Use Accuracy & Trajectory AnalysisMid28m→52Red Teaming & Adversarial Testing: Automated & Manual ApproachesMid25m→53Red-Teaming LLM Applications with LangGraph: Graph-Based Adversarial PipelinesMid26m→54Guardrails & Content Filtering: Input/Output Safety LayersMid23m→55Hallucination Detection & Mitigation: Grounding & VerificationMid22m→56Constitutional AI & RLHF for SafetyMid24m→57Bias, Fairness & Responsible AI in ProductionMid19m→58AI Governance: Compliance, Auditing & Risk ManagementMid26m→59Interpretability & Explainability: Mechanistic & Feature-Level AnalysisMid26m→60Observability: Tracing, Logging & LLM MonitoringMid23m→61Online Evaluation: Production Sampling, Drift Detection & Feedback LoopsMid20m→62AI Gateways: Rate Limiting, Fallbacks & Multi-Provider RoutingMid26m→Phase 6 · Ship on Cloudflare11 lessons
Take it to production on the edge — deploy, scale, cost-control, CI/CD, and applied multimodal patterns
63Start hereEdge Deployment: On-Device Models, ONNX & WebLLMMid25m→64Cost Optimization: Token Economics, Caching & Model SelectionMid24m→65Scaling & Load Balancing: GPU Clusters, Model Parallelism & RoutingMid18m→66LLM Serving: API Design, Batching & StreamingMid23m→67Production AI Patterns: Workflows, Pipelines & ArchitectureMid27m→68CI/CD for AI: Regression Testing, Monitoring & Continuous EvalMid26m→69Search & Recommendations with LLMsMid26m→70Conversational AI: Chatbot Design, Dialogue Management & UXMid27m→71Vision-Language Models: Architecture, Training & ApplicationsMid23m→72Audio & Speech AI: ASR, TTS & Voice AgentsMid22m→73AI for Code: Copilots, Code Review & Program SynthesisMid25m→Appendix · Fine-tuning & Training6 lessons
Beyond the edge — customizing models with LoRA, RLHF and dataset curation (off the continuous-play spine)
74Start hereFine-tuning Fundamentals: Full, Freeze & Transfer LearningMid21m→75LoRA, QLoRA & Adapter Methods: Parameter-Efficient Fine-tuningMid21m→76RLHF & Preference Optimization: DPO, ORPO & PPOMid19m→77Dataset Curation: Synthetic Data, Quality Filtering & AnnotationMid23m→78Continual Learning: Catastrophic Forgetting & Knowledge RetentionMid24m→79Distillation & Model Compression: Pruning, Quantization & Student ModelsMid24m→Appendix · Other Clouds & Platforms5 lessons
Beyond the edge — AWS/Azure/GCP, containers and Kubernetes for comparison and context
80Start hereAmazon Web ServicesMid26m→81Microsoft AzureMid11m→82Google Cloud PlatformMid11m→83DockerMid11m→84KubernetesMid11m→Appendix · AWS Deep Dives9 lessons
Beyond the edge — deep-dive reference guides for core AWS services
85Start hereAWS Lambda & ServerlessMid33m→86AWS API Gateway & NetworkingMid38m→87AWS IAM & SecurityMid36m→88AWS Compute & ContainersMid39m→89AWS Storage & S3Mid36m→90AWS CI/CD & DevOpsMid43m→91AWS Architecture Patterns & Well-ArchitectedMid48m→92AWS AI/ML Services: Interview Preparation Knowledge BaseMid43m→93AWS DynamoDB & Data ServicesMid35m→Appendix · Engineering & Communication9 lessons
Beyond the edge — timeless engineering principles, LlamaIndex, and communication skills
94Start hereMicroservicesMid11m→95CI/CDMid10m→96Node.jsMid12m→97SOLID PrinciplesMid11m→98ACID PropertiesMid10m→99PostgreSQL JOINs: Inner, Outer, Cross, Self, Lateral & PerformanceMid13m→100Foreign Keys: Referential Integrity, Cascades & Migration PatternsMid19m→101LlamaIndexMid11m→102Public Speaking: Structure, Delivery, and Audience EngagementMid10m→