Building LLM-Powered Applications

Learn both applied and theoretical knowledge on working with LLMs. In this training session we will build an end-to-end application powered by LLMs. We will start with a simple interface to accept user prompts and then we will equip it with various abilities such as: interaction with our running application, extraction of free-form text into data structures, recognizing user intent, and finally a research agent that will help us make informed decisions. This will be done in three sessions each starting with a brief theoretical introduction and then a deep dive to build desired features.

Theory:

• What are LLMs and how do they work
• Embeddings and semantic token representations
• Tools and MCPs
• Context & Prompt Engineering
• LLM Agents
• The core building blocks of Agentic Applications
• Evaluation and why it is essential for scaling LLM powered apps

Practice:

• Basic chat interface with streaming
• Receiving user commands
• Adding our first tool to the LLM
• Recognizing user intent to make informed decisions
• Allowing our LLM agent to interact with our application
• Extracting data from free-form text or websites
• Creating a research agent
• Basic evaluation system for our data extraction workflow and research agent

Prerequisites: A laptop with Elixir and Postgres. Basic knowledge of Elixir/Phoenix is a plus.