Recommended prerequisites
You can succeed without these, but expect to reference outside material often if you skip them. See Essential for what you must have first, and Prerequisites for the tier overview.
On this page
Software architecture patterns
1.5A/1.5B on orchestration patterns is, underneath, asking which software architecture fits a given multi-agent problem. Recognizing the patterns first makes that chapter feel like applying something you already know, not memorizing a new taxonomy.
Monolith vs. microservices, and loose coupling
Can you explain the difference between a monolithic and a microservices architecture, and what “loose coupling” buys you?
- Software Architecture in Practice by Len Bass, chapters 1–3
Pub-sub messaging and event-driven design
Can you describe publish-subscribe messaging and identify when it’s the right choice over a direct call?
- System Design Primer · free, community-maintained reference covering pub-sub, load balancing, and replication
Architectural trade-offs in practice
For a system with 200 agents coordinating warehouse operations, would you reach for centralized or decentralized orchestration, and why? Real case studies sharpen this instinct faster than definitions do.
- Read a real architecture case study (Uber, Netflix, or AWS engineering blogs) and identify which patterns above it actually uses
Database fundamentals
1.4 (memory systems) requires understanding persistence, and Part 6 (RAG and knowledge integration) runs on vector databases specifically — 6.2A assumes you already know SQL, indexing, and ACID before it explains why vector databases relax some of those guarantees.
SQL basics
Can you write a SELECT query with a WHERE clause and a JOIN?
- SQL Tutorial · W3Schools, ~3–4 hours, free
Indexing and query performance
Do you understand what a database index does and why it speeds up some queries but slows down writes?
- Designing Data-Intensive Applications by Martin Kleppmann, chapters 1–3
ACID vs. eventual consistency
Can you name the four ACID properties, and explain why a distributed system might deliberately give one of them up?
- Same source as above; practice by setting up PostgreSQL locally and running a transaction that violates isolation on purpose
SQL vs. NoSQL vs. vector databases
Can you compare a relational database (PostgreSQL), a document store (MongoDB), and a vector database (Pinecone) on what each is actually optimized for?
- Covered from the RAG side in 6.2A: Vector Database Selection
Kubernetes fundamentals
Part 4 (production deployment and scaling) runs on Kubernetes throughout, particularly 4.3 (container orchestration). If you’re deploying to a managed Kubernetes service (EKS, GKE, AKS), this moves from recommended to essential.
Pods, services, and deployments
Can you explain what a pod is, why it’s the smallest deployable unit, and how a service exposes a set of pods to the network?
Video
A natural next step after the Docker video in Essential > Docker — same host channel style, picks up where containers leave off.
kubectl and resource limits
Can you use kubectl to view pods, services, and logs, and explain what a CPU/memory resource
request vs. limit does?
- Practice: deploy a simple application to Minikube, scale its replica count, and trigger a rolling update
Natural language processing (NLP) basics
LLMs are NLP models underneath. Part 5’s advanced reasoning patterns build on transformer mechanics, and Part 7’s optimization techniques (quantization, KV cache) target transformer components directly. This tier goes one level deeper than Essential > LLM Fundamentals — read that first.
Tokenization
Can you explain what tokenization does, why a word can become multiple tokens, and why that matters for cost and context-window budgeting?
Video
Hands-on and code-level — build the mechanics yourself rather than take them on faith. For the attention mechanism and full transformer architecture, see Essential > LLM Fundamentals rather than duplicating those videos here.
Why transformers replaced RNNs
Do you know what made transformer models different from — and generally better than — the recurrent networks (RNNs) that came before them?
- Attention Is All You Need — read at least the introduction and the architecture diagram; this is the paper that made the case
Common NLP task types
Can you name and distinguish a few standard NLP task shapes — classification, generation, question answering — and place “agent tool-calling” among them?
- Hugging Face NLP Course, chapters 1–2, free
Continue to Beneficial prerequisites, back to Essential, or to the tier overview.