Introduction to AI Agents

This course provides an introduction to the world of agentic AI. It explores the core concepts of how AI agents operate and function, while simultaneously showing the code needed to create your own. The course was designed to demystify what agents are, introducing important concepts and frameworks at a beginner level.

The primary focus of this course was giving an introduction to AI agents and to cover concepts and frameworks of how they work The learning objectives were:

• To gain an understanding of what an AI agent is and what they are utilised.
• To build a foundational knowledge of important concepts within agentic AI, such as LLM, RAG & memory.
• To learn how AI agents plan, think, and improve, especially in a multi-agent framework.
• To gain insight into how agents are created and to analyse the code to build your own .

Given this course was primarily conceptual, no independent hands-on project was built; however, the module provided an informative introduction to agentic AI in the form of 10 bite-sized videos. Each video was hosted and led by Korey Stegared-Pace, a leader in cloud computing and agentic AI at Microsoft. Every episode tackled an essential question or topic within AI development, navigating through the lifecycle from introduction to deployment. The videos followed a consistent structure: introducing a core pattern, providing foundational knowledge, and then diving into a live coding environment to demonstrate the concept in operation. Complex ideas were continuously reinforced through diagrams and visual presentations to make the information highly digestible.

The course did an exceptional job of covering multiple fundamental concepts within the world of Agentic AI.

The first three modules focused heavily on answering the question of "what is an AI agent?", exploring the differences between generative AI (like ChatGPT and general chatbots) and agentic systems. This part of the course concentrated on explaining how AI agents operate and introducing the idea of agentic frameworks like Semantic Kernel and AutoGen. Lastly, it looked into agentic design and the principles of designing efficient AI agents, focusing on architecture, use cases, and building a persona. Throughout these modules, core concepts such as LLMs, frameworks, and long & short-term memory were discussed explicitly. Ultimately, this section focused on building a strong foundation in agentic AI, showcasing how agents are designed and utilized.

Further modules focused on expanding the scope of use of agentic AI and introducing the idea of function calls to utilize external tools and separately designed mechanisms. These modules explored the idea of architecture between not just the agent, but also databases, tools, and functions. Terms that were discussed included Retrieval-Augmented Generation (RAG) and APIs. Lastly, the middle modules covered the idea of building effective AI agents, focusing on shifting from concepts to code and implementing best practices to ensure optimal performance.

The final modules looked at the advanced stages of development for agentic AI, along with the next step for AI agents through the use of multi-system agent design, design patterns, and exploring how AI agents think and reason. This highlighted the step-by-step process of planning—from creating sub-tasks and checklists, to using APIs and tools to optimise the entire process. Additionally, the final modules leaned into the question, "how do we improve AI agents?", focusing completely on evaluations, being people-oriented, and improving through continuous learning. Finally, the course covered how to deploy AI agents into production, addressing crucial considerations like hosting, scale, and security to wrap up the development lifecycle with a fully functional AI agent built throughout the course.

To complete this course it involved critical thinking & structured note taking, The tools I used:

HOVER OVER AN ASSET

Throughout this course, I developed a strong understanding of the sheer power of Agentic AI and how it will reshape the landscape of computer science. This module did an excellent job of tracing the full agentic lifecycle, from initial creation to final production. The balance of code and concept made the entire course simple to understand while still providing key information and conveying important principles. I personally enjoyed the exploration of multi-agent systems and the architecture behind AI agents. The structure and style of the videos, along with the experienced presenter, made this course incredibly informative and engaging. Ultimately, grasping the fundamentals of agentic AI helps me stand out in today's world and gives me a distinct competitive edge.