Agentic AI: Developing AI Agents Course

Lesson 1 • Agent Taxonomies and Classifications

Surveys simple reflex, model-based, goal-based, utility-based, and learning agents. Enables learners to select the right agent class for a given problem.

Lesson 2 • Large Language Models as Agent Backbones

Explains how LLMs serve as reasoning engines within agentic pipelines and their key capabilities and limitations. Connects LLM fundamentals to agent design decisions.

Lesson 3 • Agentic AI vs. Traditional Automation

Contrasts rule-based automation, RPA, and agentic AI across flexibility, adaptability, and failure modes. Clarifies when agentic approaches add genuine value.

Lesson 4 • What Makes AI Agentic

Defines agency in AI systems by contrasting reactive models with goal-directed, autonomous behaviour. Grounds the chapter's technical vocabulary in concrete behavioural distinctions.

Lesson 1 • Self-Reflection and Plan Revision

Explores reflexion-style loops where agents critique and revise their own outputs iteratively. Prepares students to build agents that improve performance without human intervention.

Lesson 2 • Planning Algorithms for Agents

Surveys classical planning, tree-of-thought search, and heuristic approaches applicable to LLM-based agents. Equips students to choose planning strategies matched to task complexity.

Lesson 3 • Chain-of-Thought Reasoning

Introduces step-by-step reasoning elicitation through prompting and its effect on agent decision quality. Establishes the baseline reasoning technique extended in later sections.

Lesson 4 • Task Decomposition Strategies

Teaches hierarchical and sequential decomposition of complex goals into executable subtasks. Directly enables agents to handle long-horizon objectives introduced in later chapters.

Lesson 1 • Tool-Use Fundamentals

Defines what constitutes a tool in an agentic context and how agents select and invoke tools via function calling. Establishes the integration model used throughout the chapter.

Lesson 2 • Building a Custom Tool Library

Guides students through designing, documenting, and registering reusable tools for agent frameworks. Produces a portable tool library applicable to projects in subsequent chapters.

Lesson 3 • Search and Knowledge Retrieval Tools

Integrates web search and document retrieval into agent pipelines to ground responses in current information. Bridges tool use with the retrieval concepts expanded in Chapter 5.

Lesson 4 • Connecting Agents to APIs

Covers REST API integration, authentication patterns, and error handling within agent tool loops. Enables students to extend agents with any web-accessible service.

Lesson 1 • Types of Agent Memory

Distinguishes in-context, external, episodic, and semantic memory and their trade-offs in agent design. Provides the taxonomy applied in all subsequent memory sections.

Lesson 2 • Vector Databases for Agent Memory

Covers embedding-based storage and similarity search as a scalable long-term memory solution for agents. Connects to retrieval-augmented generation patterns introduced in Chapter 5.

Lesson 3 • Context Window Management

Addresses strategies for fitting relevant information into finite context windows without losing critical state. Directly impacts agent reliability in long-running tasks.

Lesson 4 • Memory Consistency and Conflict Resolution

Addresses contradictions between memory sources and strategies for maintaining a coherent agent world model. Prepares students to build reliable agents in dynamic environments.

Lesson 1 • Advanced Retrieval Strategies

Covers hybrid search, re-ranking, HyDE, and multi-query retrieval to improve recall and precision. Directly improves agent answer quality on complex knowledge tasks.

Lesson 2 • Evaluating RAG Pipeline Quality

Applies faithfulness, relevance, and context recall metrics to assess and improve RAG performance. Closes the chapter with a systematic quality assurance workflow.

Lesson 3 • Agentic RAG Patterns

Introduces iterative, self-querying, and corrective RAG patterns where agents decide when and how to retrieve. Extends basic RAG into dynamic, agent-driven retrieval workflows.

Lesson 4 • RAG Architecture Fundamentals

Explains the retrieve-then-generate pipeline, its components, and how it integrates with agent reasoning loops. Establishes the architectural baseline for the chapter.

Lesson 1 • Conflict Resolution and Consensus

Addresses disagreements between agents through voting, arbitration, and confidence-weighted merging strategies. Produces robust multi-agent outputs even under agent disagreement.

Lesson 2 • Multi-Agent Architecture Patterns

Surveys hierarchical, peer-to-peer, and market-based multi-agent topologies and their suitability for different tasks. Provides the structural vocabulary for the chapter.

Lesson 3 • Orchestrator and Subagent Design

Defines the orchestrator's role in task routing, delegation, and result aggregation across specialised subagents. Teaches the division of responsibility central to scalable multi-agent systems.

Lesson 4 • Orchestration Frameworks in Practice

Applies leading open-source orchestration frameworks to build and run a multi-agent workflow end to end. Translates architectural concepts into working code and deployable pipelines.

Lesson 1 • Prompt Injection and Adversarial Inputs

Explains how malicious inputs can hijack agent reasoning and actions, and covers detection and defence strategies. Directly protects agents deployed in open or untrusted environments.

Lesson 2 • Alignment Techniques for Agents

Applies constitutional AI, RLHF concepts, and value specification methods to keep agents aligned with intended goals. Connects alignment theory to practical agent configuration.

Lesson 3 • Ethical Frameworks and Responsible Deployment

Applies fairness, transparency, and accountability principles to agent design decisions and deployment policies. Prepares students to meet organisational and societal expectations for responsible AI.

Lesson 4 • Human-in-the-Loop Design

Defines checkpoints, approval gates, and escalation paths that keep humans appropriately in control of agent actions. Balances autonomy with accountability in production deployments.

Lesson 1 • Testing Strategies for Agent Systems

Applies unit, integration, simulation, and adversarial testing to validate agent behaviour before and after deployment. Reduces production incidents through systematic pre-release validation.

Lesson 2 • Observability and Monitoring

Implements tracing, logging, and alerting for agent pipelines to detect failures and performance regressions. Enables continuous visibility into agent behaviour in production.

Lesson 3 • Agent Deployment Architectures

Covers containerisation, serverless, and microservice patterns for hosting agent workloads in production. Establishes the infrastructure foundation for all deployment topics in the chapter.

Lesson 4 • Continuous Improvement and Iteration

Establishes feedback loops using production logs, user signals, and evaluation datasets to drive ongoing agent improvement. Closes the course with a sustainable agent lifecycle management practice.