| created | 2025-09-09 | ||
|---|---|---|---|
| last_update |
|
||
| reading_time | 14 |
Cloud Native Architecture Engineering Playbook & Benchmark Reports — Battle-tested engineering know-how from production environments
Engineering Playbook is a comprehensive collection of cloud native architecture engineering practices accumulated from production environments. It covers Amazon EKS infrastructure optimization, Agentic AI platform engineering, AIDLC/AgenticOps methodology, hybrid infrastructure, security governance, industry-specific solution patterns, and quantitative benchmark results.
Each technical domain provides implementation guides alongside measurable performance data to support data-driven architecture decisions.
Live Documentation: https://devfloor9.github.io/engineering-playbook/
| Endpoint | Purpose |
|---|---|
| llms.txt | llmstxt.org index — all technical docs with links and summaries |
| llms-full.txt | Full-text merge of all technical docs (single file) |
| llm-wiki/manifest.json | LLM Wiki manifest — per-doc metadata (slug, domain, tags, related docs, markdown URL) |
| llm-wiki/index.md | LLM Wiki index — per-page clean markdown files grouped by domain |
The LLM Wiki mirrors the 7 technical domains (industry demos excluded) as clean per-page markdown — MDX/JSX stripped, links normalized — so agents can fetch exactly the pages they need without HTML parsing. Each doc page also exposes <link rel="alternate" type="text/markdown"> pointing to its markdown source.
End-to-end guide for building enterprise Agentic AI platforms on EKS.
- Design & Architecture: Platform foundations (architecture, challenges), platform selection (SageMaker / AgentCore / EKS decision framework, AWS-native, EKS open architecture), advanced patterns (self-improving agent loop, knowledge feature store, semantic caching)
- Model Serving & Inference: GPU infrastructure (EKS GPU node strategy, GPU resource management, NVIDIA GPU stack, AWS Neuron stack), inference frameworks (vLLM, llm-d, MoE serving, NeMo, HyperPod inference operator), inference optimization (KV Cache-aware routing, disaggregated serving, LMCache, cache-hit strategy), inference routing
- Operations & MLOps: Agent monitoring & observability (Langfuse, LLMOps tooling), RAGAS evaluation, Kagent Kubernetes agents, AI Gateway guardrails, compliance framework, domain customization, Milvus vector database
- Reference Architecture: Inference Gateway setup & routing (kgateway + agentgateway + Bifrost), custom model deployment & pipeline, model lifecycle (continuous training), SageMaker-EKS integration, open-weight model deployment, OpenClaw AI Gateway
Production-grade guides for Amazon EKS infrastructure optimization.
- Networking & Performance: Cilium ENI, Gateway API migration, CoreDNS tuning, East-West traffic optimization
- Control Plane & Scaling: Large-scale cluster scaling strategies, cross-cluster object replication
- Resource & Cost: Karpenter autoscaling, resource optimization, cost management
- Operations & Reliability: GitOps (Argo CD), node monitoring, EKS debugging & resiliency, Pod health lifecycle
- Security & Authentication: EKS API server authentication/authorization, Pod Identity, IRSA
AI Development Lifecycle framework and Agentic AI operational feedback loops.
- AIDLC Framework: Reliability dual-axis (Ontology × Harness) based AI development lifecycle
- AgenticOps: OpenTelemetry observability, CloudWatch AI integration, DevOps Guru predictive operations
On-premises GPU infrastructure and cloud native platform integration.
- EKS Hybrid Nodes adoption guide
- SR-IOV with DGX H200
- File storage strategies
- Harbor registry integration
Enterprise security governance practices.
- Identity-First Security (EKS Pod Identity)
- GuardDuty Extended Threat Detection
- Kyverno policy management
- Supply Chain Security
- Default namespace incident analysis
Red Hat OpenShift on AWS installation, security, and compliance guide.
Industry-validated PoC patterns and runnable demo assets — focused on what value to show customers and why, complementing the how-to engineering guides in other sections.
- Retail: Five reference PoCs built on a common stack (Knowledge Graph on Neptune, hybrid search with OpenSearch + Cohere, persona switcher, Agentic AI on Bedrock + AgentCore, Bedrock Guardrails)
- LG H&H Marketing Innovation — 3-BU (Beauty + HDB + Refreshment) integrated marketing, 8 scenarios with 4 external signal sources
- AMWAY Direct Selling — ABO/IBO multi-level org visualization, subscription lifetime, direct-selling compliance (11 scenarios)
- Shinkong Mitsukoshi Luxury — department-store VIP membership, foreign-tourist tax-free recommendation, luxury brand SOV (11 scenarios)
- Momo eCommerce — live-commerce attribution, 24h delivery SLA, recommendation diversity (11 scenarios)
- Uni-President BU Integration — cross-BU OPENPOINT journey, own-SKU sell-through, cold-chain SLA (11 scenarios)
Quantitative benchmarks for infrastructure, AI/ML, and hybrid environments.
- Networking: CNI performance comparison (VPC CNI vs Cilium), Gateway API implementation benchmarks
- AI/ML Inference: AI/ML workload analysis, AgentCore vs EKS self-hosted inference, Dynamo inference benchmark
- Infrastructure & Operations: Infrastructure performance, hybrid infrastructure, security operations metrics
| Area | Technologies |
|---|---|
| Container Orchestration | Amazon EKS, EKS Auto Mode, Karpenter, MNG + DRA |
| Networking | Cilium, Gateway API, CoreDNS, kgateway |
| AI/ML Serving | vLLM, SGLang, llm-d, NVIDIA Dynamo, NeMo Framework, Amazon Bedrock |
| AI Gateway | kgateway + agentgateway, Bifrost, LiteLLM, OpenClaw |
| GPU Management | NVIDIA GPU Operator, DCGM, DRA, MIG, KAI Scheduler, NIXL |
| MLOps | Kubeflow, MLflow, KServe, SageMaker |
| Vector DB | Milvus |
| Observability | Prometheus, Grafana, Langfuse, Hubble, OpenTelemetry |
| Cost Tracking | Bifrost (infra-level), Langfuse (app-level) |
| GitOps | Argo CD |
| Security | Kyverno, GuardDuty, EKS Pod Identity |
| AI Agent | Kagent, MCP (Model Context Protocol) |
| Evaluation | RAGAS |
docs/
├── agentic-ai-platform/ # Agentic AI Platform
│ ├── design-architecture/ # Foundations, platform selection, advanced patterns
│ ├── model-serving/ # GPU infra, inference frameworks, optimization, routing
│ ├── operations-mlops/ # Observability, governance, data infrastructure
│ └── reference-architecture/ # Inference gateway, model lifecycle, integrations
├── eks-best-practices/ # EKS Best Practices
│ ├── networking-performance/ # Networking (Cilium, Gateway API, CoreDNS)
│ ├── control-plane-scaling/ # Control Plane Scaling
│ ├── resource-cost/ # Resource & Cost Optimization
│ ├── operations-reliability/ # Operations & Reliability
│ └── security-authn/ # Security & Authentication
├── aidlc/ # AIDLC Framework
│ ├── methodology/ # Methodology (DDD integration, ontology × harness)
│ ├── toolchain/ # Tools & implementation
│ ├── enterprise/ # Enterprise adoption
│ └── operations/ # AgenticOps
├── hybrid-infrastructure/ # Hybrid Infrastructure
├── security-governance/ # Security & Governance
├── rosa/ # ROSA (OpenShift on AWS)
├── industry-solutions/ # Industry Solutions
│ └── retail/ # Retail PoCs (LG H&H, AMWAY, Shinkong, Momo, Uni-President)
└── benchmarks/ # Benchmark Reports
Presentation materials are available at /slides:
- Agentic AI Platform — Full platform overview (86 slides)
- Inference & Model Performance Optimization — EKS architecture for LLM inference optimization
# Install dependencies
npm install
# Start dev server
npm start
# Production build
npm run buildRequires Node.js >=20.0 and Docusaurus 3.9.2
Issues, PRs, and feedback are all welcome. See GitHub Issues for details.
Content in this project is available under the MIT License.