Level: Beginner to Intermediate
Prerequisite: None
Building a demo of an AI agent is easy, but engineering a reliable, goal-oriented system that can function autonomously is a significant challenge. This full-day, hands-on workshop moves beyond the hype to teach the core cognitive architectures—reasoning, planning, and memory—required to build dependable agentic workflows. We begin with first principles, exploring agentic design patterns and strategies for effective context management to ensure agents remain grounded and efficient.
Throughout the day, you will bridge the gap between low-code prototyping and Python-based engineering, exploring various modes of tool usage and integration. We will evolve traditional RAG into "agentic RAG," enabling systems that don't just find information but solve complex problems. The course culminates in a deep dive into AgentOps, utilizing a specialized framework to ensure observability, reliability, and safety. By moving from "vibe checks" to rigorous evaluations, you will learn how to transform experimental agents into high-performance systems ready for real-world application.
You Will Learn
- The principles of agentic architecture, including reasoning loops and memory management
- How to implement standard agentic design patterns like Reflection and Tool Use
- Strategies for building agents using both no-code/low-code tools and Python code
- Multi-agent orchestration patterns, including manager-worker and peer-to-peer delegation
- Advanced context management techniques and various modes of tool integration
- Methods for evolving standard RAG into agentic retrieval systems for complex queries
- The core pillars of AgentOps: observability, tracing, and performance monitoring
- How to apply a rigorous evaluation framework to move agents into a production state
- Best practices for responsible AI and safety guardrails within autonomous workflows
Geared To
AI engineers and data scientists
- Software engineers
- Technical leads
- Technology architects
- Innovation officers
- Product managers overseeing AI/ML initiatives
Prerequisites
A basic understanding of LLMs and prompt engineering is recommended
Familiarity with Python is helpful
Laptop Setup
Students must bring their laptop to class
Note on Corporate Laptops:
This course requires the installation of software, access to Google Colab and other web services, as well as the ability to download data files.
If your corporate laptop blocks such activities, we recommend you bring and use a personal device instead.
System Requirements:
Windows PC or Mac with:
- The open source tool VS Code, installed
- A modern browser
- Access to the free Google Colab service
- The Google Colab VS Code extension, installed
Setup:
Instructions will be emailed to registrants prior to the event to prepare your laptop before the conference.
There is no time allotted in class for laptop preparation.
* Enrollment is limited to 34 attendees.