Agentic & Autonomous AI

Goal: Understand the technology behind AI agents before building them.

• 1.1Introduction to Artificial Intelligence, machine learning and generative AI
• 1.2What are Large Language Models (LLMs)? How do they work?
• 1.3Tokens, context windows, temperature and model limitations
• 1.4Prompt engineering: instructions, context, examples and structured outputs
• 1.5Introduction to Claude, Gemini, ChatGPT and other AI models
• 1.6Hallucinations, reasoning limitations and responsible AI use

Goal: Understand what makes an AI system agentic and when agents are useful.

• 2.1What is Agentic AI?
• 2.2AI assistants vs. AI agents vs. traditional automation
• 2.3Anatomy of an agent: model, instructions, tools, state and memory
• 2.4Goals, planning, task decomposition and decision-making
• 2.5The agent loop: observe, decide, act and evaluate
• 2.6Single-agent vs. multi-agent architectures
• 2.7When to use an agent, a workflow or a combination of both

Goal: Connect AI agents to external applications and services.

• 3.1Workflow automation fundamentals
• 3.2Introduction to n8n and visual workflow building
• 3.3APIs, HTTP requests, JSON and authentication
• 3.4Webhooks, triggers and application integrations
• 3.5Function calling and tool calling
• 3.6Connecting agents to email, spreadsheets, calendars and databases
• 3.7Handling API failures, invalid inputs and rate limits

Connecting an LLM to an API does not automatically make the whole system a reliable agent. You need to control what the model can call and validate the results.

Goal: Move beyond simple chatbots to agents that perform tasks.

• 4.1Agent instructions, goals and constraints
• 4.2Tool selection and tool execution
• 4.3Planning and breaking complex tasks into smaller steps
• 4.4Short-term state and conversation memory
• 4.5Persistent memory and storing information between sessions
• 4.6Retrieval-Augmented Generation (RAG) and knowledge bases
• 4.7Human approval, retries, stopping conditions and error handling

Goal: Design systems where multiple agents or specialized components work together.

• 5.1Sequential, routing and parallel agent patterns
• 5.2Manager-agent and specialist-agent architectures
• 5.3Agent handoffs and shared state
• 5.4Multi-agent collaboration and task delegation
• 5.5Tool permissions and coordination
• 5.6Avoiding infinite loops, duplicated tasks and conflicting outputs
• 5.7Comparing single-agent and multi-agent performance

Multi-agent systems are not always better. A single agent or a deterministic workflow may be cheaper, faster and easier to maintain.

Goal: Understand the code-level foundations for building more flexible, customizable agents.

• 6.1Python fundamentals for AI development
• 6.2Variables, functions, loops, dictionaries and modules
• 6.3Environment variables, API keys and package installation
• 6.4Calling LLM APIs and parsing structured responses
• 6.5Implementing tools and tool-calling loops
• 6.6Introduction to an agent framework such as LangGraph or the OpenAI Agents SDK
• 6.7State, control flow, persistence and resumable tasks

You will build the same task twice: first in n8n, then in code. That shows you the underlying architecture instead of leaving you dependent on a visual builder.

Goal: Make your agents dependable and safe enough for real use.

• 7.1Prompt injection and untrusted external content
• 7.2Protecting API keys, credentials and customer information
• 7.3Least-privilege access and tool permissions
• 7.4Human-in-the-loop approvals for sensitive actions
• 7.5Testing, evaluation datasets and output quality
• 7.6Logging, monitoring, latency and cost tracking
• 7.7Deployment, webhooks, environment configuration and uptime
• 7.8Failure recovery, versioning and maintenance

Learn to tell a prototype that works in a demo from a production system that can be operated responsibly.

Goal: Build a complete solution to a real problem and demonstrate your technical and business skills.

• 8.1Select one project
• 8.2Define its scope
• 8.3Build it
• 8.4Test it
• 8.5Deploy it
• 8.6Present its value