AI Systems Are Software Systems
Why production AI requires the same engineering discipline as any distributed system, and how this wiki covers the full stack of AI …
Intermediate
GitHub Actions Security: Risks, Exploits, and Hardening
A comprehensive guide to GitHub Actions security vulnerabilities, common exploit patterns, and how to audit and harden your CI/CD pipelines …
Build a Code-Based Video: Programmatic Video Production with Remotion
A step-by-step guide to creating professional demo and explainer videos entirely in code using Remotion, React, and AI-assisted development. …
Everything as Code: Treating All Artifacts as Software
The principle of defining infrastructure, configuration, documentation, policy, video, and design as version-controlled code artifacts - and …
Version Control Fundamentals and Git
AWS Well-Architected AI/ML Lens - Applying Best Practices to Machine Learning
The AWS ML Lens extends the Well-Architected Framework to cover ML lifecycle phases, ML pipeline automation, model security, inference …
Caching Patterns for AI Applications
Semantic caching, Anthropic prompt caching, response caching, and embedding caching for AI applications. Cost savings analysis and …
Hardware Constraints for AI Systems
CPU vs GPU, VRAM limits, memory bandwidth, and how hardware choices determine what AI models you can run and at what cost.
Hybrid Cloud
What hybrid cloud is, why it matters for AI workloads with data gravity and compliance constraints, and AWS hybrid options including FSx for …
Programming Languages for AI - Python, TypeScript, HCL
A practical guide to the three languages used across a modern AI stack: Python for agents and models, TypeScript for frontends and video …
Prompt Engineering for Enterprise AI Applications
Practical prompt engineering patterns for production AI systems: system prompts, few-shot examples, chain-of-thought, structured output, …
Retry and Backoff Patterns for AI Services
Exponential backoff with jitter, retry budgets, and idempotency patterns for production AI systems. Why AI services require different retry …
Sorting and Search Algorithms for AI Pipelines
How sorting and search algorithms underpin AI pipeline design: complexity trade-offs, partial sorting for top-k selection, tiered analysis …
The Well-Architected Framework - Why Every Cloud Provider Has One
What the Well-Architected Framework is, its origins at AWS, how Azure and GCP adopted it, its six pillars, and why it matters especially for …
Tiered Analysis Pattern - Progressive Depth for AI Processing
Apply cheap analysis first, score results, then apply expensive analysis only to candidates that pass a threshold. Reduces AI API costs by …
AI Architecture Patterns - From Monolith to Multi-Agent
How AI system architecture evolves from monolithic single-model deployments through microservices to collaborative multi-agent systems, with …
AI Deployment Models - SaaS, PaaS, IaaS, and Serverless
How the four cloud deployment models apply to AI workloads: when to use managed models, platform endpoints, GPU instances, or serverless …
Amazon CloudWatch - Monitoring and Observability for AI
Using Amazon CloudWatch for AI workloads: custom metrics for LLM cost and token usage, alarms for model quality, log insights for inference …
Amazon EventBridge - Event-Driven AI Orchestration
Using Amazon EventBridge to connect AWS AI services, trigger pipelines from S3 events, and build loosely coupled multi-step workflows.
Amazon OpenSearch Service - Search and Analytics for AI
Using Amazon OpenSearch Service for vector search, full-text search, and log analytics in AI-powered applications.
AWS Elemental MediaConvert - Video Processing at Scale
Using AWS Elemental MediaConvert for transcoding, format conversion, and video processing in AI media pipelines.
Blue-Green Deployment
What blue-green deployment is, how it works, why it matters for zero-downtime AI model updates, and how it compares to canary and rolling …
Blue-Green Deployment for AI Services
Zero-downtime model updates using blue-green deployment: how it works, AWS implementation with Lambda aliases and SageMaker variants, and …
Canary Deployment
What canary deployment is, how gradual traffic shifting works, which metrics to watch, and how to configure automatic rollback triggers for …
Canary Deployment for AI Models
Gradual traffic shifting to new model versions: how to implement canary deployments with Lambda weighted aliases and SageMaker production …
CI/CD Pipelines for AI Projects
Building reliable CI/CD pipelines for AI projects: model artifact management, automated evaluation gates, GitHub Actions workflows, and …
Circuit Breaker Pattern
What the circuit breaker pattern is, why AI services need it for handling model timeouts and rate limits, and how to implement it with AWS …
Circuit Breaker Pattern for AI Services
Handling model failures gracefully in production AI systems: fallback strategies, degraded mode operation, retry with backoff, and …
Event Sourcing
What event sourcing is, why it matters for AI audit trails and pipeline replay, its relationship to CQRS, and when to apply it in AI …
Event Storming for AI Use Case Discovery
How to run an Event Storming workshop specifically for discovering AI automation opportunities: domain events, commands, policies, and …
Feature Flags
What feature flags are, how they enable safe AI model rollouts, A/B testing, and instant rollback - and the tools available for implementing …
Feature Flags for AI Model Deployment
Using feature flags to safely roll out AI model changes: A/B testing models, canary deployments, gradual traffic shifting, and instant …
Impact Mapping for AI Projects
Applying the Why-Who-How-What Impact Mapping framework to AI projects: grounding AI initiatives in measurable business outcomes and avoiding …
Infrastructure as Code for AI Projects
Why IaC matters for AI reproducibility, multi-environment consistency, and cost tracking. Terraform and CDK patterns for Bedrock agents, …
Model Context Protocol (MCP) - Universal Tool Interface for AI Agents
What the Model Context Protocol is, how it enables AI agents to use tools through a standard interface, and server/client architecture.
Model Drift and Data Drift
What drift is, the three types (data, concept, prediction), how to detect them using SageMaker Model Monitor, and when to trigger model …
Model Versioning and Artifact Management
Why model versioning matters and how to implement it: S3 for artifacts, Git for configuration, SageMaker Model Registry, Bedrock model …
Observability
What observability means, the three pillars of logs, metrics, and traces, and why AI systems need specialized observability for token costs, …
Observability for AI Systems - Logs, Metrics, Traces
Applying the three pillars of observability to AI workloads: CloudWatch for metrics and alarms, Langfuse for LLM tracing, OpenTelemetry for …
Open Practice Library for AI Projects - Discovery to Delivery
Applying Open Practice Library practices to AI: Event Storming for AI use case discovery, Impact Mapping for AI value, User Story Mapping …
Property-Based Testing
What property-based testing is, why it is ideal for AI systems that cannot be tested with exact-output assertions, and the tools available …
Pydantic AI - Type-Safe Agent Development
Using Pydantic AI to build AI agents with validated inputs and outputs, Bedrock backend support, and Python type annotations.
Strands Agents - AWS-Native Agent SDK
What Strands Agents is, how it differs from CrewAI and LangGraph, and when to use it for AWS-hosted agent applications.
Testing AI Systems - Unit Tests to Production Monitoring
A practical testing strategy for AI systems: property-based testing, integration testing with mocked models, evaluation frameworks, and …
The Shared Responsibility Model for AI on AWS
How AWS shared responsibility applies to AI and ML workloads: data, model, and infrastructure responsibilities across Bedrock and SageMaker.
AI Cost Optimization Patterns
Model selection by task, caching strategies, batch vs real-time processing, and tiered inference with Haiku, Sonnet, and Opus.
AI for Citizen Services - Modernizing Government Interactions
Chatbot-based citizen inquiries, form pre-filling, status tracking, and multilingual support for government agencies.
AI for Recruitment and Talent Screening
How AI assists recruitment teams with resume screening, candidate matching, and interview scheduling - with guidance on bias mitigation and …
AI Guardrails - Safety and Compliance Controls
What AI guardrails are, the types of controls they enforce, how to implement them in enterprise applications, and Amazon Bedrock Guardrails …
AI Readiness Assessment - Is Your Organization Ready?
A five-dimension self-assessment to understand where your organization stands before committing to an AI program.
AI Transcription with Accurate Speaker Attribution
How to achieve production-quality multi-speaker transcription with speaker diarization, using AWS Transcribe and Bedrock post-processing.
AI-Powered Accessibility for Broadcasters and Media
Automated subtitle generation, audio descriptions, sign language overlay detection, and WCAG compliance checking for broadcast and media …
Amazon Bedrock vs Azure OpenAI - Which to Choose?
A practical comparison of Amazon Bedrock and Azure OpenAI Service for enterprise AI deployments, covering model selection, pricing, …
Amazon Cognito - User Authentication for AI Apps
Amazon Cognito User Pools and Identity Pools: JWT token structure and expiry, MFA options, SAML/OIDC federation, Lambda triggers, rate …
Amazon Comprehend - NLP at Scale
Sentiment analysis, entity extraction, topic modeling, and language detection with Amazon Comprehend. When to use Comprehend vs Bedrock for …
Amazon Rekognition - Image and Video Analysis
What Rekognition does, which features work well in enterprise applications, accuracy considerations, pricing, and common integration …
Amazon SageMaker vs Bedrock - Build vs Buy
When to use SageMaker for custom ML versus Bedrock for managed foundation models - a practical comparison for enterprise AI teams.
Amazon Textract - Document Data Extraction
A reference guide to Amazon Textract: OCR capabilities, table and form extraction, query-based extraction, and integration patterns for …
Amazon Transcribe - Speech-to-Text for Enterprise
Amazon Transcribe capabilities, accuracy characteristics, pricing, and the integration patterns that work well for enterprise transcription …
Automated Content Metadata and Tagging with AI
Auto-tagging video and audio content, scene classification, topic extraction, and SEO metadata generation for media libraries.
AWS AI Services vs Azure AI - Complete Comparison
A service-by-service map of AWS AI and ML services to their Azure AI equivalents, covering language models, speech, vision, and MLOps.
AWS AI Services vs Google Cloud AI - Complete Comparison
A service-by-service map of AWS AI and ML services to their Google Cloud equivalents, covering language models, speech, vision, and MLOps.
AWS Amplify - Full-Stack App Development
Using AWS Amplify to deploy front-end applications, host static sites, and connect to AWS AI backends.
AWS Lambda for AI Pipelines
Serverless inference, event-driven processing, and integration patterns with Bedrock, SageMaker, and Step Functions. Cost optimization for …
AWS Step Functions - Workflow Orchestration for AI Pipelines
How Step Functions orchestrates multi-step AI workflows, handles retries and errors, and integrates with other AWS services - with practical …
AWS Step Functions vs Lambda Chains for AI Orchestration
When to use state machines vs direct invocation for AI workflows. Error handling, retry patterns, cost comparison, and visibility …
Building Enterprise AI Chatbots That Actually Help
Practical guidance for building customer-facing AI chatbots that deliver real value - architecture, knowledge base design, escalation …
Conference-Driven Development: Building and Presenting AI Systems in Public
How the discipline of preparing conference talks produces better AI prototypes, clarifies system design, and accelerates learning. Covers …
Context Window Management Patterns
Summarization, sliding window, retrieval-augmented, and hierarchical context patterns for handling conversations and documents that exceed …
CrewAI - Multi-Agent Orchestration Framework
What CrewAI is, how it models multi-agent systems as crews with roles and tasks, integration with LLM backends, and when to use it versus …
CrewAI vs Strands Agents - Multi-Agent Framework Comparison
Architecture differences, AWS integration, and decision criteria for choosing between CrewAI and Strands Agents for multi-agent AI systems.
Custom ML Models vs Foundation Models - When to Build vs Buy
SageMaker custom training vs Bedrock foundation models. Data requirements, cost, accuracy trade-offs, and maintenance burden.
Data Pipeline Patterns for AI/ML Workloads
Practical patterns for building reliable data pipelines that feed AI and ML systems - ingestion, transformation, feature engineering, and …
Data Preparation for AI Projects - A Practical Guide
How to prepare data for AI projects: assessing what you have, cleaning and normalizing it, building evaluation datasets, and setting up …
Embeddings - Vector Representations for AI Search
What embeddings are, how they enable semantic search, which embedding models to use, and how to choose vector database infrastructure.
Event-Driven Architecture for AI
What event-driven architecture is, how S3 triggers, EventBridge, and Step Functions patterns enable scalable AI pipelines.
Fine-Tuning vs Prompt Engineering vs RAG
The three main approaches to customizing LLM behavior for specific use cases - when each is appropriate and how they compare.
From AI Idea to Working Prototype in 3 Workshops
A structured three-workshop methodology that takes an organization from AI curiosity to a validated, buildable prototype with stakeholder …
How to Facilitate an AI Workshop - A Practitioner's Guide
Preparation, agenda design, stakeholder management, use case brainstorming techniques, prioritization exercises, and gap management between …
Intelligent Document Processing with AI
A practical architecture for extracting structured data from invoices, contracts, and forms - combining OCR, classification, and LLM-based …
Langfuse - LLM Observability and Tracing
Using Langfuse to trace LLM calls, evaluate outputs, and monitor AI application quality in production.
LangGraph - Stateful AI Agent Graphs
How LangGraph models AI agent workflows as stateful graphs, enabling cyclic execution, human-in-the-loop, and complex multi-step agent …
LlamaIndex - RAG and Agent Framework
Using LlamaIndex for retrieval-augmented generation, data connectors, and agent workflows, with Bedrock and OpenSearch integration.
Multi-Agent AI Systems - When One Model Is Not Enough
A practical introduction to multi-agent AI architectures: when to use them, how they work, and which frameworks are production-ready.
Multi-Agent Systems
Definition, architecture patterns, and frameworks for multi-agent AI systems - and the signals that indicate a single-agent approach is no …
Prompt Engineering Patterns for Enterprise Applications
Proven prompt patterns for enterprise AI applications: structured output, chain-of-thought, few-shot examples, guardrails, and system prompt …
RAG vs Fine-Tuning - When to Use Each
A practical framework for deciding between retrieval augmented generation and fine-tuning to customize LLM behavior for enterprise …
Remotion - Programmatic Video Creation with React
Using Remotion to generate videos programmatically from React components, with Lambda rendering for scalable AI-driven video production.
Remotion vs FFmpeg - Video Processing Approaches
When to use Remotion (React-based programmatic video) vs FFmpeg (command-line video processing) for AI video pipelines.
Terraform - Infrastructure as Code for AI Projects
Using Terraform to provision and manage AWS infrastructure for AI projects: modular design, state management, and multi-environment …
Terraform vs AWS CDK - Which IaC Tool to Choose
When to use Terraform vs AWS CDK for AI project infrastructure: pros, cons, and decision criteria for each tool.
The Use Case Scoring Framework - From 57 Ideas to 3 Prototypes
A structured WSJF-inspired scoring methodology to cut through workshop noise and identify the AI use cases worth building first.
Vector Database
What vector databases are, how they enable semantic search, popular options including Pinecone, Weaviate, and pgvector, and when to use …
Why Your AI Output Sounds Generic - And How to Fix It With Your Own Data
The difference between prompting and grounding. Five stages from zero context to production-ready assets. The Personal Inference Pack …
Well-Architected Framework
The cloud architecture review methodology used by AWS, Azure, and Google Cloud to evaluate workloads against proven best practices across …