Open teaching · CC BY-SA 4.0
Agentic AI: foundations, architectures, and applications
A complete, open course on agentic AI. Designed for professionals and advanced students building systems with LLMs and agents. Available in English and Spanish under CC BY-SA 4.0.
The agent loop
The agent loop
Every agent architecture rides on the same cycle: perceive, reason, act, observe. The architectures you'll meet later are variations on this loop.
The agent loop
Perception
01 / 04
Start here
Enter the course content
Fifteen interactive lectures with generated infographics, simulators, and hands-on exercises. No downloads, no PDFs, everything inside the web.
Programme
15 weeks, 4 blocks
Click any week to open its summary, core concepts, and the centrepiece infographic that anchors it.
Introduction to AI Agents
Foundational definitions of AI agents as systems that perceive, reason, and act. Historical evolution from BDI agents and reinforcement learning to LLM-based agents, with the Russell & Norvig taxonomy mapped onto modern agentic systems.
Open lectureCore concepts
- Perception-action loop
- Agent autonomy
- Russell-Norvig taxonomy
Centrepiece infographic
The perception-action loop
Labs
8 hands-on labs with real code
Each lab is self-contained and finishable in a two-hour session. Starter kits ship as runnable Python templates.
Your First LLM API Calls
Set up the Python environment and explore generation parameters end-to-end.
Building a ReAct Agent from Scratch
Implement the Thought-Action-Observation loop without any framework.
Advanced Tool Integration
Use native function calling with five diverse tools and an intelligent router.
Building Agent Memory
Implement and compare buffer, summarisation, and vector-based memory architectures.
Building a RAG Pipeline
Full RAG pipeline with chunking, retrieval, re-ranking, and citations.
Multi-Agent Debate System
Two agents debate, an orchestrator picks a winner, and we score argument quality.
Building an Evaluation Framework
Define metrics, build a test suite, run LLM-as-judge with statistical analysis.
Safety and Red-Teaming
Run adversarial tests; implement input guards, output filters, and action constraints.
Adopt the course
Everything is open
Teaching it? Reach out and I'll support you. Studying it? The labs guide more than any video.