Agentic AI

My notes from the Agentic AI course by DeepLearning.AI (taught by Andrew Ng), which I took from October 28 to November 5, 2025. The course covers how to build agentic workflows: breaking tasks into steps, the four main design patterns (reflection, tool use, planning and multi-agent collaboration), and how to evaluate and improve these systems in practice.

📁 Documentation structureCopy Anchor Link

Each module has its lecture notes, and the labs are converted from the executed Jupyter notebooks.

The notebooks are also on GitHub, with the extra files (data, helper modules) needed to run them locally:

MuzzammilShah/Agentic-AI

Code implementations of the graded and ungraded labs.

00

🔎 Sample output: research agentCopy Anchor Link

Module 1 ends with a deep-research agent that was already built for the course. I asked it about Manchester City under Pep Guardiola, and it planned the research, searched the web, drafted and reviewed the text, and produced the report below on its own. It is reproduced unedited, so its facts and citations were not checked.

✍️ Final notesCopy Anchor Link

That's the end of this! The course was absolutely amazing, took me about a week to wrap it up, and I'd 100% recommend y'all watch it as well.

You'll also notice this kind of felt like introductory content, but I think it covers the essentials and gives us a good idea of how to first identify an agentic system and how we can actually label our work as a 'qualified agent' (since there are many projects out there that claim they are actually "agents," haha).

You can find the course on the official deeplearning.ai website and all of the images in these notes are credited to them. I hold no ownership of them and am only using them for purely educational and informational purposes (and am not commercializing this content in any way).

Thank you for reading, now we move on to the next one! Happy learning :)