Learning Paths

Four ways through this site, depending on why you're here. Pick the one that matches your situation — none of them require reading everything in order.

🎯 SE/SA Entering AI (the flagship path)

For someone like the author: an SE/SA building AI fluency from the ground up, in the order that gets you fluent fastest.

1. How LLMs Actually Work (no math) → The Four-Layer Map → AI Vocabulary for SAs → LangGraph in 10 Minutes
2. Managed API vs Self-Host → Do We Even Need an Agent? — the two build-vs-buy calls that come up first
3. Scoping an AI POC — how an engagement actually gets scoped
4. Lab 01 · First LLM App → Lab 02 · Production RAG — the first real, hands-on proof
5. Explaining a Hallucination — the talk track that closes the loop

🔥 In a Live AI Deal Right Now

For when you need an answer in the next five minutes, not a curriculum. This is also START-HERE.md's "live deal" section — same path, linked from wherever you land first.

1. START-HERE — the fastest-orientation entry point
2. Match your situation to a talk track — Explaining a Hallucination if the model just made something up
3. Match it to a decision frame — Managed API vs Self-Host, The Real Cost of a RAG System, Do We Even Need an Agent?, How Do We Know It Is Good Enough?, or What Will This Cost at Scale?
4. Scoping an AI POC — the recovery section covers what to do when a live demo goes sideways

🛠️ Get Your Hands Dirty (build credibility)

For proving you can build the real thing, not just talk about it. Every lab is $0 and local-first.

1. Lab 01 · First LLM App — the app loop every later lab builds on
2. Lab 02 · Production RAG — hybrid retrieval, rerank, grounding
3. Lab 04 · Eval Harness — evals early, on purpose, before agents
4. Lab 03 · Agent System — hub-and-spoke + one MCP tool
5. Lab 05 · Serving & Cost — measure instead of guess
6. Lab 06 · Observability (Phase 5) — trace what you built
7. Lab 07 · Capstone (Phase 5) — assemble everything above into one agent

🔀 From DevOps to AI (the crossover)

For infrastructure/DevOps engineers who already have the ops instincts and need the AI-specific layer on top, not a rebuild from scratch.

1. The MLOps↔LLMOps Bridge — LLMOps extends MLOps, it doesn't replace it
2. Reference Architectures → Guardrails and Governance — where the new layer sits, and what governs it
3. DevOps Studio — the classic infra practice this path bridges from, sibling site in the trilogy

🔗 The Trilogy

This site is the third of three, one through-line — a Solutions Architect who makes hard technical domains understandable:

• Solutions Playbook — how engagements actually run
• DevOps Studio — the infrastructure layer
• AI Engineering Studio — this site, the newest and least-understood domain

Each stands alone; read together, they cover the ground a Solutions Architect is expected to hold.