Agentic Design Patterns

This folder holds the design-pattern track of the guide: the catalog of recurring shapes that show up once you start assembling language models into systems that actually do things, and — more to the point — how to talk about them in a system-design interview.

Two things live here.

The Agentic Design Patterns Interview Playbook is the main event: all twenty-one patterns, each written as a complete interview preparation unit, one page per pattern so you can jump straight to the one you’re drilling. Start at the introduction, then the vocabulary primer and the universal answer skeleton before the patterns themselves. For every pattern you get a realistic scenario the way an interviewer would actually phrase it, a plain-English explanation of what the pattern is and what problem it solves, a walkthrough of how the design works including where it breaks, and then a long model answer written the way you would say it out loud in a fifteen-minute design discussion — clarifying questions, tradeoffs, failure modes, evaluation, cost, and a whiteboard diagram you could sketch. It closes with a pattern picker mapping common scenarios to the patterns they call for, and four composed mega-scenarios showing how the patterns stack in real answers. Part 1 ends with a self-quiz and its answer key — attempt the questions aloud before opening the answers.

It assumes no production-systems vocabulary. Terms like idempotency, tail latency, backpressure, and circuit breakers are defined in plain language the first time they appear, because half of sounding senior in an interview is using those words comfortably and the other half is knowing what they mean when the interviewer says them first.

Agentic_Design_Patterns.pdf is Antonio Gulli’s book, the source catalog the playbook is built around. The playbook is a companion, not a replacement — the book carries the code examples and framework-level detail (LangChain, LangGraph, Google ADK, CrewAI) that the playbook deliberately sets aside to focus on explanation and delivery. Read the book for implementation; read the playbook to rehearse.

There is also a good hour-long video summary of the same catalog at https://www.youtube.com/watch?v=e2zIr_2JMbE. The playbook’s final appendix maps its timestamps to the chapters here so you can watch a section, pause, and immediately practice the matching answer.

How this relates to the rest of the guide

The numbered chapters (01_ through 12_) are about evaluating agents — measuring whether they work, catching regressions, monitoring them in production. This folder is about architecting them. The two meet constantly: Pattern 19 in the playbook is the interview-level view of evaluation, and it points into those chapters for the depth. Similarly, AGENT_ENGINEERING_FOUNDATIONS.md at the repo root covers hands-on building — frameworks, MCP, RAG, memory, cost engineering — with runnable code.

A reasonable path through all three: read the playbook to learn the shapes and rehearse talking about them, read the foundations doc to build one for real, and read the numbered chapters to prove it works.