Core Concepts

• What Large Language Models (LLMs) and Small Language Models (SLMs) are, how they work, and when to use each
• The transformer architecture, attention, context windows, and Mixture-of-Experts (MoE)
• How generative AI systems differ from traditional software and ML systems
• What AI agents are, how they reason, and how they act using tools

 

Key Techniques

• Prompting strategies and structured prompts
• Retrieval-Augmented Generation (RAG) for grounding models in documents and databases; naive RAG, two-stage RAG, metadata filtering, embeddings, vector databases
• Tool use and function calling to extend model capabilities
• Fine-tuning models to improve performance and reduce cost
• Designing and orchestrating agent workflows and multi-step tasks

 

Agentic AI & Automation

• Fundamentals of building AI agents that:
  • Gather data
  • Make decisions
  • Execute actions
• Multi-agent collaboration and orchestration patterns
• Event-driven agents triggered by external systems (e.g. GitHub, APIs)
• Architectural considerations for scaling agent-based systems

 

Tools & Frameworks

• LLMs from OpenAI and other providers
• Locally hosted SLMs
• Vector databases and embedding models
• Agent frameworks such as CrewAI, Agno and others
• The Model Context Protocol (MCP) for secure, standardized access to external tools and services

 

Practical Outcomes

By the end of the workshop, participants will be able to:

• Integrate LLMs into applications and business processes
• Build RAG-powered systems over documents and data sources
• Design and implement AI agents that automate real tasks
• Understand the trade-offs between different models, tools, and architectures
• Leave with working examples and sample code applicable to real-world projects

 

This 2-day, hands-on workshop provides a practical introduction to Generative AI and Agentic AI, emphasizing not just the use of AI tools, but the understanding, design, and architecture of modern AI-powered systems deployed in real-world applications.

 

DAY 1

Language Models (LLMs) – Large Language Models (LLMs) are a boon to software development because they permit apps to do things that were impossible just a few short years ago. They are also the chief enabling technology behind AI agents.

Get up close and personal with LLMs from OpenAI and others, learn what differentiates one from another, and learn how to leverage these models to write software that’s more intelligent than ever before. Also become acquainted with Small Language Models (SLMs) that can run locally and do much of what their LLM cousins can do without incurring per-token costs for input and output.

 

Retrieval-Augmented Generation (RAG) – A popular use case for LLMs in industry today is putting them over internal documents to make information in those documents easily discoverable. Retrieval-augmented generation, or RAG, is a technique for extending a language model’s knowledge base to include information contained in PDFs, DOCX files, and other documents. Moreover, it’s a means for erecting guardrails around a language model so the answers it gives are truthful and restricted to information in the documents you provide.

Learn what RAG is, how it works, and how to build sophisticated RAG pipelines of your own by combining popular language models with vector databases, text embeddings, and cross encoders. Covers naive RAG, two-stage RAG, metadata filtering, and more.

 

How Large Language Models Work – In 2017, a landmark paper entitled “Attention is All You Need” turned the deep-learning world on its head and laid the foundation for today’s Large Language Models. In the paper, we learned about the transformer architecture and its unparalleled ability to understand, and even to generate, human language. This section takes a deep dive into natural-language processing, starting with simple word embeddings and text classifiers, moving to Recurrent Neural Networks (RNNs), and ultimately landing on transformers.

You’ll come away understanding what a transformer is, how it works, and what happens as words flow through a transformer and are converted from static embeddings with low information density into rich, context-aware embeddings that capture the meaning of the words around them. You’ll also learn what context windows are, why they exist, and why GPUs are so crucial to a transformer’s operation. Finally, you’ll learn about the Mixture-of-Experts (MoE) architecture and why virtually all of today’s foundation models use it.

 

DAY 2

Making LLMs Smarter – LLMs generate text. Unaided, they can’t access the Internet, perform basic math, or access databases. But equip them with tools and they can do all this and more. Learn how to supercharge LLMs with function calling and tool use, how to put LLMs over documents and databases, how to lend them superpowers (including the ability to produce colorful charts and graphs from your data) by leveraging their ability to generate code,  how to use fine-tuning to increase performance while reducing cost, and more.

 

The Model Context Protocol (MCP) – The Model Context Protocol is an open standard that permits LLMs to access external services and data sources. It has been called the “USB-C of AI applications,” and it might be the most important development in AI since language models themselves.

Learn what MCP is, why it’s important, how to write MCP clients and MCP servers, how to secure MCP servers, how to decide if your company should expose services through MCP, and how to leverage MCP to build truly intelligent AI-enabled applications.

 

AI Agents – Agentic frameworks such as CrewAI and Agno simplify the process of building AI agents that work alone or as part of a team. Agents can be triggered to act in response to external stimuli such as PRs submitted to GitHub. Agents can also orchestrate complex workflows, freeing users from the tyranny of prescriptive UIs. It’s now entirely possible to build an interface that accepts a command (typed or spoken) such as “Find all the nonstop flights from Atlanta to New York on June 1st that cost less than $800 and identify the three with the most unsold seats in first class” — or just about any other command you could dream up.

Learn how to build agents and put them to work in your business and see some jaw-dropping examples of the tasks they can perform. Also learn how to architect agentic applications so that they scale. (Hint: Many do not, especially when agents working together to accomplish a task are required to share data.)