The systems layer after prompt engineering

Build AI Systems That Think in Graphs

Move beyond isolated prompts and fragile agent loops. Learn to connect agents, models, code, tools, memory, verification, and human decisions into reliable systems you can inspect and ship.

$6.93single course · lifetime access — or Full Access membership at $19.63/mo, first 30 days free
6 modules · 10 tools · 10 expert perspectives · 3 projects
Planner Researcher A Researcher B Skeptic Reducer Verifier repair ↺
a work graph: fan-out → reduce → verify → bounded repair
Core positioning

The Progression

Prompt engineering determines what a model is asked to do. Loop engineering enables an agent to repeatedly reason, act, observe, and adjust. Graph engineering determines how agents, models, tools, code, memory, verification, and human judgment work together as an explicit system.

Prompt

One instruction

One instruction produces one response. Powerful, but stateless and blind to consequence.

Loop

One agent

One agent repeatedly reasons, acts, and observes. Capable — until it becomes impossible to inspect and control.

Graph

One system

Specialized agents, tools, code, memory, verification, and humans coordinate through explicit structure.

Prompts generate responses. Loops perform tasks. Graphs operate systems.

Graph Engineering Mastery is not a course about making more agents. It is a course about making intelligent systems:

ExplicitInspectableReliableCost-boundedRecoverablePersistentHuman-governedSmall enough to understand
Three connected reference layers

The 10–10–10 Learning Ecosystem

10 Tools

Real frameworks

Students use real frameworks and infrastructure in controlled laboratories — not toy sandboxes.

10 Expert Perspectives

Reference thinkers

Each module introduces the thinking of important builders, researchers, and educators shaping the field.

10 External Courses

Curated continuation

A curated continuation path instead of an overwhelming directory of links — opened in a separate “Continue Learning” area.

The purpose is not to master every framework. The purpose is to understand the underlying architecture well enough to move between frameworks.

Learn the architecture, not just one framework

The Ten Essential Graph-Engineering Tools

You will learn where each tool is strong, where it creates unnecessary complexity, and how to select the smallest appropriate stack.

LangGraph

Primary workflow-graph framework

Typed state, conditional routing, parallel execution, cycles, checkpoints, persistence, human interrupts, long-running stateful agents.

Modules 1 · 2 · 4 · 6

Pydantic Graph

Type-safe graph & state-machine lab

Explicit node contracts, typed inputs and outputs, return-type-defined edges, finite-state-machine thinking, runtime detection of invalid transitions.

Modules 1 · 3

Google ADK

Enterprise & multi-language workflow lab

Sequential, parallel, and loop workflows; multi-agent composition; evaluation and deployment across Python, TypeScript, Go, Java, and Kotlin.

Modules 2 · 5

Microsoft Agent Framework

Type-safe business-process lab

Executors and edges, graph-based workflows, checkpointing, concurrent execution, human-in-the-loop processes. Successor direction to AutoGen, now in maintenance mode.

Modules 2 · 5 · 6

OpenAI Agents SDK

Handoff, guardrail & tracing lab

A deliberately small abstraction layer: agents, runners, tools, handoffs, guardrails, sessions, run traces, usage and cost tracking.

Modules 2 · 5 · 6

CrewAI

Role-based team & automation lab

Specialist roles, Crews for autonomous collaboration, Flows for controlled event-driven orchestration, memory, knowledge, guardrails, observability.

Modules 2 · 5

LlamaIndex Workflows

Data-intensive & event-driven lab

Agentic retrieval, document workflows, router agents, multi-document research, structured outputs, event-driven workflow steps.

Modules 3 · 5

DSPy

Graph optimization & evaluation lab

Declarative LM modules, multi-stage programs, metrics, automated instruction optimization, improving nodes against measurable outcomes.

Modules 4 · 6

Neo4j + GraphRAG

Persistent knowledge-graph lab

Entities and relationships, Cypher, provenance, knowledge extraction, graph traversal, GraphRAG, agent memory, explainable retrieval.

Modules 3 · 5 · capstone

LangSmith

Tracing, evaluation & observability lab

Execution traces, node-level inspection, dataset-based evaluation, failure diagnosis, latency analysis, production monitoring. A graph that cannot be observed is not ready for production.

Modules 4 · 6
Six core modules

The Curriculum

Every module is organized around one central question, two expert lenses, real tool laboratories, and one project output.

0

Orientation: Prompt, Loop, Workflow or Graph?

Outcome: students learn not to use a graph automatically.

The minimum-necessary-architecture rule. Drawing the system before coding. First exercise: classify ten problems as prompt, function, loop, workflow, graph, or human process. Personal Learning Graph activation.

Expert lenses · Andrew Ng, Jeremy Howard
1

Node and Edge Engineering

Central question: what work is actually dependent?

Nodes as bounded responsibilities. Edges as real data dependencies. Input/output contracts, structured state, false dependencies, explicit stopping, finite-state-machine thinking, Mermaid architecture diagrams.

Tool labs · Pydantic Graph, LangGraphExpert lenses · Samuel Colvin, Harrison Chase
Project output: convert a fragile linear agent into a typed four-to-six-node graph.
2

Orchestration and Multi-Agent Topology

Central question: what can run in parallel, and who decides what happens next?

Sequential edges, conditional routing, fan-out, fan-in, joins and reducers, supervisor–worker graphs, specialist routers, handoffs, dynamic versus predetermined execution, human escalation.

Tool labs · LangGraph, Google ADK, Microsoft Agent Framework, OpenAI Agents SDK, CrewAIExpert lenses · Chi Wang, João Moura
Project output: build a multi-perspective research graph with parallel researchers, a reducer, and a conditional reviewer.
3

State, Memory and Knowledge Graphs

Central question: what should exist after the current run ends?

Working state, checkpoints, session and durable memory, entity extraction, canonicalization, typed relationships, provenance, confidence, contradictions, GraphRAG, memory consolidation.

Extract → Resolve → Assemble → Verify → Persist → Repeat

Tool labs · LlamaIndex, Neo4j, LangGraph persistence, OpenAI Agents SDK sessionsExpert lenses · Jerry Liu, Emil Eifrem
Project output: turn an unstructured document collection into a source-grounded knowledge graph an agent can traverse.
4

Verification, Evaluation and Improvement

Central question: how does the system know whether it is becoming better?

Read-only critics, verification gates, evidence checking, bounded repair cycles, node- and graph-level evaluation, dataset creation, metric design, automated optimization, regression testing, cost-quality tradeoffs.

Tool labs · DSPy, LangSmith, Pydantic EvalsExpert lenses · Omar Khattab, Matei Zaharia
Project output: create an evaluation suite and optimize one graph node against a defined metric without degrading the rest of the graph.
5

Cross-Framework Graph Studio

Central question: which implementation best fits this graph?

Compare LangGraph, Pydantic Graph, Google ADK, Microsoft Agent Framework, OpenAI Agents SDK, CrewAI, and LlamaIndex Workflows across control, type safety, state, persistence, parallelism, human interaction, observability, knowledge integration, deployment, and learning curve. Students do not choose a framework because it is fashionable — they choose it because its execution model matches the problem.

Project output: implement the same small architecture in two different frameworks and explain the tradeoffs.
6

Production and Capstone

Central question: can this graph survive real users, failures, and changing requirements?

Tracing, replay, checkpoint recovery, idempotency, retries, timeouts, spend caps, permission boundaries, tool isolation, prompt-injection defense, state migration, versioning, model routing, caching, human approval, graph simplification.

Tool labs · all ten tools available — students use only the minimum stack their project requires
Final principle: the best graph is not the graph with the most nodes. It is the smallest graph that reliably protects the outcome.
Reuse proven structure

Nine Reusable Graph Patterns

Reuse proven structures instead of reinventing every workflow. Each pattern ships as an editable template.

1 · The Chain

Use when every step truly depends on the previous step.

A → B → C

2 · Fan-Out / Fan-In

Use when independent workers can operate simultaneously.

       ┌→ B ─┐
A ─────┼→ C ─┼→ E
       └→ D ─┘

3 · Conditional Router

Use when state determines which specialist should act.

Input → Router ─┬→ Specialist A
                ├→ Specialist B
                └→ Human

4 · The Diamond

Split work, perform it independently, then merge the results.

    ┌→ Left  ─┐
In ─┤         ├→ Merge
    └→ Right ─┘

5 · Generator & Verifier

One node produces. Another independently evaluates.

Generate → Verify ─┬→ Accept
                   └→ Repair

6 · Human Approval Gate

Pause before consequential actions.

Draft → Review ─┬→ Approve
                ├→ Edit
                └→ Reject

7 · Supervisor & Specialists

A coordinator delegates bounded tasks without performing every task itself.

Supervisor ─┬→ Worker A
            ├→ Worker B
            └→ Worker C

8 · Persistent Knowledge Loop

Extract, resolve, verify, and preserve knowledge across runs.

Extract → Resolve → Assemble
   ↑                    │
Repeat ← Persist ← Verify

9 · Multi-Speed Improvement Graph

Separate fast operational adjustments from slower strategic, safety, and audit decisions.

Fast loop:  tune → retry
Slow loop:  audit → govern
Three shippable projects

Three Capstone Paths

You will not simply watch someone build them. You will draw, run, inspect, repair, and ship them yourself.

Capstone A

Verified Research & Publishing Graph

  • Planner & parallel researchers
  • Source collection & reducer
  • Draft generator
  • Skeptical critic & evidence verifier
  • Human approval
  • Publishing output
Capstone B

Persistent Personal Knowledge Graph

  • File ingestion & entity extraction
  • Canonicalization & typed edges
  • Provenance & confidence
  • Contradiction handling
  • Graph retrieval & memory updates
  • Agent router
Capstone C

Production Multi-Agent Organization

  • Specialist agents & deterministic tools
  • Conditional routing & parallel work
  • Shared state & persistent memory
  • Verification & bounded repairs
  • Human escalation & observability
  • Cost controls & architecture docs
Learn from the builders shaping the field

Ten Expert Perspectives

These experts form the course's intellectual reference map — GraphEngineerings.com independently synthesizes their public work into one architecture-first learning path.

Harrison Chase

Stateful orchestration & explicit agent architecture

Co-founder and CEO of LangChain, closely associated with LangGraph and the movement from simple model chains toward persistent, inspectable agent workflows.

Andrew Ng

Agentic design patterns & progressive AI education

Founder of DeepLearning.AI. His agentic curriculum emphasizes reflection, tool use, planning, and multi-agent workflows built first from principles.

Jeremy Howard

Practical-first learning

Founding researcher at fast.ai. His teaching lets students build useful systems early, then progressively exposes the deeper theory.

Chi Wang

Multi-agent conversation & collaboration

Senior author of the original AutoGen research: customizable agents combining models, tools, code, and human input through flexible interaction patterns.

Jerry Liu

Agents working over data & documents

Co-founder and CEO of LlamaIndex; teaches agentic retrieval patterns in which agents route, investigate, and reason across document collections.

João Moura

Role-based multi-agent teams

Founder of CrewAI; teaches the organization of specialized agents using roles, memory, tools, collaboration, guardrails, and controlled flows.

Samuel Colvin

Type safety & trustworthy Python systems

Creator of Pydantic; leads the ecosystem containing PydanticAI, Pydantic Graph, evaluation, durable-execution integrations, and Logfire observability.

Omar Khattab

Programming & optimizing LM systems

Led the creation of DSPy, which models LM applications as modular computational graphs optimized against measurable outcomes.

Matei Zaharia

Systems infrastructure & evaluation

His work includes DSPy and related systems for optimizing AI programs, prompts, code, and semantic computation.

Emil Eifrem

Relationships as first-class data

CEO and co-founder of Neo4j; has spent his career building and advocating graph databases and the property-graph model.

These experts are not presented as formal faculty or endorsers of GraphEngineerings.com. Their inclusion does not imply partnership, instruction, affiliation, or endorsement.
Every lesson, six layers

The Lesson Experience

Lessons take 8–15 minutes. No long passive lecture is required before you begin building.

1

Core Idea

One concise architectural principle.

2

Expert Lens

How one relevant expert or project approaches the problem.

3

Tool Lens

A comparison of how two frameworks represent the same idea.

4

Build

A runnable micro-project.

5

Verify

Automated structural and output evaluation.

6

Connect

The result is added to your Personal Learning Graph.

Understand → Draw → Build → Run → Inspect → Verify → Explain
Continue learning

Ten Curated External Learning Paths

A curated continuation path, not an overwhelming directory of links — each chosen for what it does best.

Practical Deep Learning for Coders — fast.ai

Practical-first AI foundations; build working applications early.

Agentic AI — DeepLearning.AI

Reflection, tool use, planning, and multi-agent patterns from first principles.

AI Agents in LangGraph — DeepLearning.AI

The bridge from a Python agent loop to a flow-based LangGraph application.

Introduction to LangGraph — LangChain Academy

The official structured introduction to state, nodes, edges, memory, and orchestration.

Multi-Agent Systems with CrewAI — DeepLearning.AI

Plan, build, debug, observe, and deploy role-based multi-agent systems.

Building Agentic RAG with LlamaIndex — DeepLearning.AI

Routers, tools, research agents, and multi-document reasoning.

AI Agentic Design Patterns with AutoGen — DeepLearning.AI

Design patterns for collaboration and reflection — studied as a historical framework.

Hugging Face AI Agents Course

Broad free path: fundamentals, smolagents, LlamaIndex, LangGraph, Agentic RAG, certification.

Hugging Face Context Course

Agent skills, MCP, plugins, subagents, lifecycle hooks, and a minimal agent harness.

Generative AI and GraphRAG — Neo4j GraphAcademy

Graph fundamentals, GraphRAG, knowledge construction, retrieval, and production graph apps.

Issued for demonstrated work

Graph Engineering Practitioner

The credential is issued for a working, explainable graph — not for merely watching lessons.

Certification requires demonstrated competence in:

  1. Node and edge design
  2. Typed state
  3. Conditional routing
  4. Parallel execution
  5. Verification
  6. Bounded repair
  7. Persistent memory
  8. Evaluation
  9. Observability
  10. Human governance
Simple by design

Pricing

Try for Free
30 days
full access trial · no card required
  • Everything in Full Access
  • Prompt-or-Graph diagnostic
  • All six modules
  • Graph canvas & starter patterns
  • Personal Learning Graph
  • Cancel-free: it simply ends
Try for Free — 30 Days
Single Course
$6.93
one-time · lifetime access to Graph Engineering Mastery V1
  • Six complete modules
  • Ten tool laboratories
  • Nine reusable graph patterns
  • Three capstone paths
  • Downloadable templates
  • Practitioner certificate path
  • Version 1 content updates
Buy the Course — $6.93
AI execution, kept honest. The single course includes starter execution credits; membership includes monthly credits. Continued AI usage runs through bring-your-own API key, local model mode, or free simulation mode — keeping prices honest without promising unlimited variable-cost model usage.
Final brand statement

The objective is a builder, not a name-collector

Graph Engineering Mastery brings together the best current ideas from agent orchestration, graph databases, context engineering, workflow systems, AI evaluation, and practical technical education. But the objective is not to produce a student who knows ten framework names. The objective is to produce a builder who can look at a complex problem and determine:

01 What deserves to be a node
02 What requires an edge
03 What can happen in parallel
04 What state must travel
05 What knowledge should persist
06 What must be verified
07 What should remain deterministic
08 What requires human authority
09 What should be measured
10 What should be removed
Draw the work. Expose the dependencies. Verify every important transition. Preserve trusted knowledge. Ship the smallest graph that works.