Agentic AI Systems

A deep, text-first course on how agentic systems are designed: the agent loop, architecture, tool calling, workflows, multi-agent coordination, memory and retrieval, and the reliability engineering that makes agents safe to ship. Every concept comes with interactive diagrams, worked examples, exercises and scenario quizzes.

What you will be able to do

• Decide when an agent is the right design, and when a workflow or plain code is better
• Architect an agent runtime: context, tools, state, planning and termination
• Build a tool-calling agent with validation, error handling and tests
• Compose sequential, parallel, routed and human-approved workflows
• Design multi-agent systems without drowning in coordination failures
• Engineer memory, retrieval and context budgets
• Make agents reliable: guardrails, retries, circuit breakers, evaluation and tracing

Modules

1. Module 1 · ~22 minIntroduction to AI AgentsWhat an agent actually is, how its loop works, and, just as important, when you should not build one.
2. Module 2 · ~28 minAgent ArchitectureThe parts of an agent runtime, how they fit together, and the design decisions, and failure modes, hiding in each one.
3. Module 3 · ~30 minBuilding Your First AgentFrom task definition to a tested, bounded tool-calling agent, with validation, error handling, logging and the mistakes nearly everyone makes the first time.
4. Module 4 · ~28 minAgentic WorkflowsSequential, parallel, routed, looping and human-approved workflows, how to compose them, pass state through them, and decide how much autonomy each step deserves.
5. Module 5 · ~30 minMulti-Agent SystemsWhen splitting work across several agents helps, the coordination topologies that work, and the new failure modes, loops, deadlocks, cascades, you take on in exchange.
6. Module 6 · ~30 minMemory, Context & KnowledgeHow agents remember, what they should forget, and how retrieval turns a stateless model into one that can work with your data, without drowning it in tokens.
7. Module 7 · ~32 minReliable & Trustworthy AgentsEngineering agents you can ship: validation, guardrails, retries, timeouts, circuit breakers, permissions, prompt-injection defence, evaluation and observability.
8. Module 8 · ~35 minCapstone: Design an Agentic SystemPut every module together: design, justify and stress-test a complete agentic system, from requirements to architecture, reliability, evaluation, cost and rollout.

Skills you will have by the end

Complete all 8 modules and these are the skills this course builds:

• Agentic AI
• AI Agent Architecture
• Tool Calling / Function Calling
• LLM Workflow Orchestration
• Multi-Agent Systems
• Retrieval-Augmented Generation (RAG)
• Context & Memory Engineering
• AI Guardrails & Reliability
• LLM Evaluation & Observability