Skip to content

GenAI Learning Curriculum

Self-paced, beginner-to-advanced coverage of the full modern GenAI stack —
from foundational language models through fully autonomous agentic systems.

The Stack

Every layer builds on the one before it:

01 LLMs → 02 Prompts → 03 RAG → 04 MCP → 05 Agents → 06 Agentic AI

Engine · Interface · Memory · Protocol · Actors · Systems

Layer What it does
LLMs The engine — a model that predicts the next token
Prompts The interface — communicating intent precisely
RAG The memory — injecting external knowledge at query time
MCP The protocol — standardizing how tools connect to any model
Agents The actors — LLMs that loop, use tools, and take multi-step action
Agentic AI The systems — multiple agents coordinating autonomously

Curriculum Topics

# Topic What You'll Learn Notes Code Labs Q&A
01 LLM Models Transformers, attention, KV cache, training, fine-tuning, GPU, production 12 — 68+
02 Prompt Engineering Zero-shot → chain-of-thought → meta-prompting, production patterns 4 — ✓
03 RAG Embeddings, chunking, retrieval, evaluation, Vertex AI, scaling 12 — 80+
04 MCP Protocol definition, components, architecture, getting started 8 — ✓
05 Agents Agent fundamentals + ADK, LangChain, LangGraph, CrewAI deep-dives 10 Agent Types ✓
06 Agentic AI Multi-agent systems, architectural & design patterns, evaluation 6 Architectures · Systems ✓

Code Labs

Hands-on practice across three dimensions — every pattern implemented in all four frameworks (ADK · LangChain · LangGraph · CrewAI) using the same task for direct side-by-side comparison.

Dimension 1 — Agent Evolution

→ 05-Agents: Code Labs → Agent Types

Build progressively more capable agents in all four frameworks:

Level What You Build
Simple ReAct loop + tools
Intermediate Memory, context tracking
Complex Planning, reflection, multi-step reasoning

Dimension 2 — Architecture Patterns

→ 06-Agentic AI: Code Labs → Architecture Patterns

Seven coordination patterns, each implemented × 4 frameworks:

Pattern Key Concept
Sequential Agents run in strict order
Parallel Agents run concurrently, results merged
Hierarchical Supervisor delegates to sub-agents
Orchestrator-Subagent Central planner + specialized workers
Pipeline Output of one agent feeds the next
Adversarial Debate Two agents argue; a judge decides
Reflexion Agent critiques and rewrites its own output

Dimension 3 — End-to-End Systems

→ 06-Agentic AI: Code Labs → Agentic Systems

Four production-style systems, each × 4 frameworks:

System What It Does
Research Assistant Web search + synthesis + structured report
Document Processor Ingest → extract → summarize → classify
Autonomous Task Planner Break goal → plan steps → execute → adapt
Code Review System Analyze → critique → suggest → validate

Framework Comparison

Framework Mental Model Best For
Google ADK Agent as a configurable object GCP-native, Vertex AI, production
LangChain Chain of callables (LCEL) Flexibility, broad ecosystem
LangGraph State machine / graph Stateful flows, cycles, branching
CrewAI Role-playing crew Multi-agent collaboration, readable code

Learning Paths

Path A — Conceptual (new to GenAI)

  1. LLM Fundamentals
  2. Transformer Architecture
  3. Attention Mechanisms
  4. Prompt Basics
  5. Core Techniques
  6. RAG Fundamentals

Path B — Interview Preparation

  1. LLM Models — all 12 notes
  2. RAG — all 12 notes
  3. Agent Fundamentals
  4. Agentic Concepts
  5. Architectural Patterns
  6. All Q&A — Knowledge Check

Path C — Hands-On Engineering

  1. Agent Types Code Lab — simple → intermediate → complex
  2. Architecture Patterns Code Lab — 7 patterns × 4 frameworks
  3. Agentic Systems Code Lab — 4 end-to-end systems
  1. LLM Fundamentals
  2. Transformer Architecture
  3. Prompt Basics → Core Techniques
  4. RAG Fundamentals → Embeddings & Vector Stores
  5. MCP — The Problem → MCP — Components
  6. Agent Fundamentals → Agent Patterns
  7. LangGraph Fundamentals ← pick one framework
  8. Agent Types Code Lab
  9. Agentic Concepts → Architectural Patterns
  10. Architecture Patterns Code Lab
  11. All Q&A — Knowledge Check

Knowledge Check

All concept review Q&A by topic:

Topic Link
All topics (single file) All Questions
LLM Models LLM Models Q&A
Prompt Engineering Prompts Q&A
RAG RAG Q&A
MCP MCP Q&A
Agents Agents Q&A
Agentic AI Agentic AI Q&A

Quick Start

Notes — no setup needed

pip install mkdocs-material
mkdocs serve
# → http://127.0.0.1:8000

Code Labs

python -m venv .venv && source .venv/bin/activate
pip install -r setup/requirements.txt
cp setup/.env.example .env
# Edit .env → set GOOGLE_API_KEY
jupyter notebook docs/05-Agents/CodeLabs/01-Agent-Types/ADK/01-simple/agent.ipynb

Found this useful? Star the repo — it helps others discover it.
★ Star sundante/sun-course-genai on GitHub