AI Superstream: AI Agents

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

Introduction – Antje Barth (5 minutes)

Antje Barth welcomes you to the AI Superstream.

Keynote: Build Tomorrow with AI Agents – Antje Barth (15 minutes)

Discover how AI agents will transform industries and reshape our digital world. In this keynote, Antje Barth offers a visionary look at the future of intelligent systems, exploring how AI-driven automation and adaptive technologies can redefine the way we build, innovate, and interact with technology.

4 Patterns of AI-Native Development and Conway's Law – Patrick Debois (30 minutes)

AI isn’t replacing developers—it’s changing where developers create value. Get a glimpse of the DevOps, GenAI, and software delivery of the future in this proactive session from DevOps pioneer Patrick Debois, coauthor of the DevOps Handbook. As software development undergoes a new phase of automation, new tools—many of them powered by agentic AI—are emerging and leading to new practices. Patrick dives into four patterns of practice to help you understand and navigate AI-native development and explores the socio-technical impact of agentic tools, highlighting how teams and organizations might change as they embrace this new dimension in artificial intelligence.

Break (5 minutes)

Agentic AI Explained: From Components to Conversations – Gabriela de Queiroz (30 minutes)

Agentic AI promises systems that don’t just respond—they reason, act, and adapt dynamically to complex tasks. But what’s actually required to build one? AI advisor Gabriela de Queiroz breaks down the building blocks behind modern agentic AI systems. You’ll learn how language models integrate with external tools through function calling, how grounding with data sources reduces hallucinations, and how memory and planning enable adaptive multistep behavior. To understand how these components come together in practice, Gabriela demonstrates a real-world agent that handles user queries, generates charts, and navigates multiturn conversations. Whether you're building your first agent or scaling one for production, you’ll get a clear, practical framework for moving from theory to implementation.

Putting Multiple AI Agent Systems in Production – Tony Kipkemboi (30 minutes)

Deploying multiple AI agents in a production environment presents unique challenges, including interagent communication, resource management, scalability, and system reliability. Tony Kipkemboi, a founding senior developer advocate at CrewAI, shares effective strategies for orchestrating multi-agent AI systems, emphasizing architectural patterns, deployment techniques, and monitoring practices. You’ll gain insights into real-world implementations, common pitfalls, and practical solutions to ensure seamless integration and optimal performance of AI agent ecosystems in production settings.

Break (5 minutes)

The Critical Need for AI Agent Management – Chris Hallenbeck (Sponsored by Boomi) (30 minutes)

The rise of unmanaged AI agents makes security and governance more important than ever. To scale agentic AI innovation responsibly, businesses need a clear strategy for managing risk, compliance, and control. When agents access enterprise systems and act autonomously, unchecked behavior can create serious vulnerabilities. Chris Hallenbeck, general manager of AI at Boomi, shares why governance must come first, not after something breaks. He introduces Boomi Agentstudio, the only full agent lifecycle management solution that empowers teams to design, govern, and orchestrate agents at scale. With centralized registration across providers, built-in observability, and powerful agent tools, Agentstudio gives organizations the control they need to unlock hyperproductivity responsibly.

This session will be followed by a 30-minute Q&A in a breakout room. Stop by if you have more questions for Chris.

Building Useful Agents with MCP – Lucas Soares (30 minutes)

Integrating AI models with various data sources often involves complex, custom solutions that are difficult to scale and maintain. The Model Context Protocol (MCP) addresses this challenge by providing a universal, open standard for connecting AI systems with data sources, replacing fragmented integrations with a single protocol. Noted machine learning engineer Lucas Soares explores practical implementations of MCP, showing how it simplifies development, enhances data accessibility, and improves the efficiency of AI applications. You’ll learn how to leverage MCP to build scalable AI systems that seamlessly interact with diverse data environments.

Break (5 minutes)

Task Complete? Says Who? Evaluating Agents Beyond Surface-Level Metrics – Erin Mikail Staples (30 minutes)

So your AI agent completed the task. But was it accurate? Did it hallucinate, take the scenic route, or spend 10 API calls doing something a user didn’t want? When building with LLM agents, traditional metrics like accuracy or completion rate just don’t cut it. With some real-world examples to illustrate, Erin Mikail Staples, senior developer experience engineer at Galileo, dives into why out-of-the-box metrics don’t scale for agent-driven applications—and how to build custom evaluation systems that actually reflect user needs and product goals.

High-Performance, Agentic AI Inference Systems: Innovations from DeepSeek, Nvidia Dynamo, vLLM, CUDA, and PyTorch – Chris Fregly (30 minutes)

AI agents are revolutionizing industries in many ways, including performance innovations in hardware. Chris Fregly, author and AI product leader, demonstrates how to capture the full capabilities of today’s hardware into highly-tuned compute engines for next-generation autonomous AI agents. Drawing on recent breakthroughs, he shows how modern AI inference systems deliver high throughput, low latency, and scalable performance to power AI agents. He also explains how codesigning software with cutting-edge hardware can address scaling challenges in the ultra-scale inference environments required by AI agents.

Closing Remarks – Antje Barth (5 minutes)

Antje Barth closes out today’s event.