What will you learn?

1. What actually makes an AI system an agent? 

What can an LLM do on its own? Where does it fall short? What changes when we give it the ability to interact with an environment, maintain state, and decide what to do next?

You’ll learn the difference between a standard LLM application, an LLM workflow, and a genuinely agentic system.

2. Why reasoning matters

An agent needs to work through problems, decide between possible actions, and reason about what should happen next.

We’ll introduce the role of reasoning LLMs and give you intuition for approaches such as Chain-of-Thought, search against verifiers, and other techniques that improve reasoning behavior. These topics sit at the heart of the course’s second lecture.  

3. Memory + tools: from talking to doing

A standalone LLM is fundamentally limited. Memory gives an agent useful context across interactions and actions. Tools let it do things outside the model—retrieve information, call APIs, run code, interact with software, and affect its environment.

4. Planning, reflection + autonomy

Once an agent can reason, remember, and use tools, it can begin tackling longer and less predictable tasks.

We’ll introduce how agents break problems into steps, sequence actions, respond to feedback, reflect on results, and adapt their plan along the way.

5. How do you know whether an agent is actually good?

A convincing final answer doesn’t necessarily mean the agent worked correctly.

We’ll introduce the different layers of agent evaluation, from outcomes to the trajectory of decisions and actions, along with reliability and safety.

6. Making sense of agent harnesses

Today’s agent harnesses and frameworks package many of these capabilities together. Once you understand the components underneath them, it becomes much easier to see what a framework is doing for you, and what tradeoffs it introduces.

Who is this masterclass for?

This session is designed for technical people who already use AI and want a clearer understanding of what comes next.

Software engineers moving into agentic AI 

You’ve started seeing agents everywhere and want to understand the architecture beneath the APIs and frameworks.

ML / AI engineers

You understand models, but want a more complete mental model of how reasoning, memory, tools, planning, and evaluation come together.

Data scientists and technical ML practitioners

You want to understand where agents fit into modern AI systems and how to reason about them beyond the buzzwords.

AI-curious technical leaders and builders

You need enough understanding of the agent stack to make better technical, product, or architecture decisions.

You’ll get the most out of the session if you already have basic familiarity with AI and LLMs. No previous agent-building experience is required.

RAGPACK.ai