Building AI that Doesn't Break - MLOps Mini Summit on Kubernetes Workflows, Durable AI Applications, and Agentic Tools

Explore three cutting-edge approaches to building resilient AI systems in this MLOps Community Mini Summit featuring industry experts discussing practical solutions for production challenges. Learn how to orchestrate Kubernetes workflows using Hera, the Python SDK for Argo Workflows, eliminating the need for complex YAML configurations while keeping business logic and orchestration in clean, testable Python code. Discover durable workflows that automatically checkpoint state and recover from failures, ensuring AI pipelines remain resilient when facing API issues, human delays, and other production obstacles. Understand why the future of agentic tooling requires process-based rather than function-based approaches, enabling agents to maintain state, recover from interruptions, and reliably complete tasks in customer-facing applications. The session includes real-world examples of AI pipeline implementations, practical demonstrations of workflow orchestration techniques, and insights into building stateful agents that can remember context and handle complex multi-step processes. Gain valuable perspectives from Elliot Gunton on simplifying workflow authoring for data scientists, Qian Li on abstractions for reliable cloud computing, and Alan Nichol on the evolution of conversational AI systems.