AI Agents & ML Architecture

Deep-dive articles on building production AI agent systems from scratch — architecture decisions, LLM orchestration patterns, adaptive learning loops, and how to keep your system provider-agnostic.

  • Building a Multi-Agent Invoice Approval Crew with CrewAI: A Step-by-Step Guide

    How I built a 4-agent invoice processing and approval workflow using CrewAI — from scaffolding the project on app.crewai.com, to defining agents and tasks in YAML, wiring tools with RAG embeddings, fixing a real platform validation error, and running the crew locally with uv.

    learning/ai-agents-ml-architects/crewai-invoice-approval-workflow

  • Building an AI Agent for Invoice Processing: Architecture Deep Dive

    How I designed a multi-layer, LLM-powered invoice automation system with an acting pipeline, investigation layer, and adaptive learning framework — without RAG, without LangChain, and without vendor lock-in.

    learning/ai-agents-ml-architects/invoice-processing-agent-architecture

  • From Plain LLM to Context Graph: The Evolution of AI Knowledge Retrieval

    A progressive deep-dive through four generations of AI knowledge architecture — from standalone LLMs, through RAG and GraphRAG, all the way to Context Graphs — with diagrams, real tools, and honest trade-offs at each stage.

    learning/ai-agents-ml-architects/from-llm-to-context-graph

  • The 16 Types of RAG: A Practical Field Guide

    A technical, no-hype walkthrough of the 16 most-cited RAG architectures — Standard, Hybrid, HyDE, Contextual, Recursive, Self-RAG, Modular, Memory-Augmented, GraphRAG, Knowledge-Enhanced, Agentic, Multi-Modal, Multi-Model, Federated, Streaming, ODQA, and Domain-Specific — with definitions, components, flow diagrams, and where to use (and avoid) each.

    learning/ai-agents-ml-architects/types-of-rag