GenArchitects

Learning paths

Not sure where to start? Pick the path that matches where you are today. Each path is a sequence: read the steps in order, and come back as new guides are added. Steps marked “coming soon” are on our writing plan.

Path 1 · Beginner

New to AI: from zero to your first AI application

For: graduates, developers, testers and analysts taking their first steps into generative and agentic AI. No machine learning background needed.

  1. Learn the vocabularyThe twelve concepts you will meet first, from tokens and context windows to RAG, agents and MCP, in plain English. Read the guide →
  2. How AI applications are builtComing soonPrompts, system prompts, structured output and tool calling: the building blocks of every LLM application.
  3. Build your first RAG applicationComing soonA step-by-step walkthrough: load documents, create embeddings, retrieve and answer with citations.
  4. Understand the trade-offsWhen to use RAG, fine-tuning or long context, and why the choice matters. Read the guide →
  5. Agents and safety basicsWhat changes when AI starts taking actions, and the guardrails every agent needs. Read the guide →
  6. Build a portfolio projectComing soonA guided project you can show in interviews: a small agent with tools, guardrails and evals.
Path 2 · Intermediate

Becoming an AI architect

For: senior engineers, tech leads and solution or integration architects who want to design and govern AI systems, not just use them.

  1. Refresh the concept mapSkim the core concepts with an architect’s eye: cost, latency, data and risk. Skim the map →
  2. Choose a knowledge strategyA decision framework for RAG, fine-tuning and long context, with a design-review checklist. Read the guide →
  3. Design guardrails and governanceLeast agency, scoped identity, risk-tiered approvals, policy layers and audit trails. Read the guide →
  4. Treat MCP as your agent integration layerWhat the stateless 2026-07-28 specification means for gateways, identity and deployment. Read the guide →
  5. Reference architecture: the enterprise AI platformComing soonHow models, retrieval, tools, gateways, identity and observability fit together.
  6. Evals, observability and costComing soonHow to measure quality, trace agents in production and keep token spend predictable.
  7. Communicate AI decisionsComing soonArchitecture decision records, risk assessments and how to explain AI trade-offs to the business.
Path 3 · Advanced

Forward deployed engineers: shipping AI in the field

For: forward deployed engineers (FDEs) and AI engineers who already build with LLMs and work directly with customers to get agents into production.

  1. Get current on MCPThe 2026-07-28 changes, a migration checklist and what to tell customers. Read the guide →
  2. Production guardrails checklistA field-ready checklist to run with customers before any agent gets write access. Use the checklist →
  3. Frame the knowledge decision with customersUse the RAG, fine-tuning and long-context framework in discovery workshops. Use the framework →
  4. Evals customers trustComing soonBuilding evaluation sets with domain experts and reporting quality in business terms.
  5. Debugging agents in productionComing soonReading traces, common failure modes and how to fix them quickly.
  6. Stay current every weekComing soonThe GenArchitects Radar newsletter: the week in generative and agentic AI.

Have a topic you want on a path? Reach out via LinkedIn.