Course Overview
The Agentic AI Hackathon challenges participants to build a sophisticated, multi-agent system for a fictional newspaper agency. The goal is to automate the article-writing process, from initial topic entry to final content moderation. Using frameworks like Semantic Kernel, participants will design and orchestrate a network of specialized AI agents that can work together to perform complex tasks.
This hackathon goes beyond simple chatbots, focusing on a multi-agent system where different AI agents collaborate to achieve complex goals. Participants will learn how to connect these agents and orchestrate their actions to create sophisticated, end-to-end solutions capable of interacting with databases, generating images, and performing other tasks. The hackathon builds on the concept of learning by doing, providing a hands-on environment where participants can apply their skills through a series of interconnected challenges.
Who should attend
- Developers and AI Enthusiasts: Individuals with a professional background in software development who are keen on exploring cutting-edge AI technologies and methodologies.
- AI Practitioners: Professionals and consultants who need to expand their knowledge from traditional machine learning to the new paradigm of agentic AI and multi-agent systems.
- IT Specialists: Professionals who manage and maintain an organization's technology infrastructure, including deploying and integrating new applications and services.
Prerequisites
- General Programming Skills: Proficiency in a programming language, preferably Python, is essential, as many AI frameworks and libraries are Python-based.
- Basic Understanding of Azure: Familiarity with Azure services, including creating and managing resources, is necessary to complete the challenges.
- Fundamental AI Concepts: Participants should have a basic grasp of what a large language model (LLM) is and the concepts of prompt engineering.
Course Objectives
By the end of this hackathon, participants will be able to:
- Design and architect a multi-agent AI system using modern frameworks like Semantic Kernel.
- Create and deploy specialized AI agents that perform specific tasks, such as content generation and image creation.
- Orchestrate agent collaboration to achieve complex, end-to-end goals.
- Integrate external tools and APIs (like weather services and web crawlers) into an agent's capabilities via an MCP Server.
- Utilize Azure services for key components, including Azure AI Search, Storage Accounts, and databases like PostgreSQL or Cosmos DB.
- Implement content moderation and safety checks by leveraging a separate agent and external moderation tools.
- Deploy AI agents as separate instances (e.g., containers) to ensure scalability and separation of concerns.
Outline: Agentic AI Hackathon - Azure (AAI-HACK-AZ)
The challenges are interconnected and built upon one another, guiding participants through the process of creating a complex, multi-agent AI system. The final product will be a complete article-writing automation solution.
- Challenge 1: Environment Setup and Agent Foundation
- In this initial challenge, participants will establish the core environment for their agentic AI system. This involves deploying the AI Foundry with relevant models and setting up foundational services like a Storage Account, Azure AI Search, or PostgreSQL or Cosmos DB. The goal is to create the first agent from this foundry, which will serve as the initial point of contact for the system.
- Challenge 2: Deploying the Agent's Toolset
- Participants will create and deploy an MCP Server with the necessary tools, such as web crawling, an article database, and a weather API. The goal is to make these tools available for the agent to use, enabling it to gather relevant sources for an article. This challenge focuses on enabling the agent to leverage external functions to extend its capabilities.
- Challenge 3: Connecting the Agents to the Tools
- Building on the previous challenge, participants will deploy the MCP Server to Azure and connect their AI Foundry to it. This step ensures the agent can access and utilize the tools deployed in Challenge 2 in a cloud environment. The focus is on creating a seamless connection between the agent and its external resources.
- Challenge 4: The Writing and Art Agents
- Participants will use Semantic Kernel to create two specialized agents: one that writes paragraphs of text and another that generates images. The writer agent will create content based on the sources gathered, while the image agent will either find existing pictures or generate new ones using models like DALL-E for specific landmarks or elements. For this challenge, participants will create and deploy a new MCP Server to host the tools needed for image database access and generation. This ensures a separation of concerns and appropriate security permissions.
- Challenge 5: The Orchestrator
- This challenge focuses on creating an orchestrator agent using the Semantic Kernel. This orchestrator will be responsible for managing the workflow: it will take the initial request, spin up instances of the writer agents, then the image agents, and finally gather all the content for the next step. This is the core of the agentic system.
- Challenge 6: Deploying to Azure
- In this challenge, participants will prepare their complete agentic AI solution for deployment. They will deploy all the agents to Azure, setting them up as separate instances, such as containers, to ensure they can be managed and scaled independently by the orchestrator.
- Challenge 7: The Moderator Agent
- In the final challenge, participants will create a new agent—the moderator. This agent will use external tools like AI Content Filtering and moderation services to review the generated content. Its purpose is to ensure the written text is ethically and factually correct before it is saved to the database. For this challenge, participants will create and deploy another MCP Server specifically for the moderation tools, maintaining separation of concerns and security policies.