What you’ll learn & do:
Scale Workflows Effortlessly
Discover how to expand compute clusters across clouds and reuse containerized tasks to slash latency and resource waste. Live Demo: Configure multi-cluster Flyte deployments and workload placement to balance cost, performance, and reliability.
Master Parallelism with MapTasks (and Beyond)
Learn when to use Flyte’s MapTasks for high-fan-out workflows. We’ll explore their power (low overhead, programmatic iteration) and their limits (Kubernetes metadata bottlenecks).
Live demo: Optimize a real-world workflow using MapTasks, then push it to its breaking point to see where Kubernetes struggles.
Break Through Kubernetes Limits with Union
Go beyond Flyte’s native capabilities with Union, a control plane that eliminates Kubernetes bottlenecks. See how it:
Auto-offloads metadata to blob storage when K8s etcd is overwhelmed.
Orchestrates multi-cloud clusters seamlessly, with execution placement you control.
Introduces Actors: reusable, hot containers that bypass Pod startup delays for rapid iteration.
Deep Dive: Refactor a failing MapTask into a Union-powered workflow using Actors and cross-cluster mapping.
About Union.ai
Union is the developer suite powering compound AI systems. We make building AI
fast, easy, and scalable.
Stay Connected
💬 Join our AI & MLOps Slack: https://slack.flyte.org/
⭐ Explore Flyte on GitHub: https://github.com/flyteorg/flyte
🌐 Learn more: https://union.ai/
