Flyte

Flyte is an open-source orchestration platform purpose-built for reproducible, scalable, infra-aware machine learning and AI workflows. It helps teams run and manage durable workloads for training, batch inference, serving, and agent loops, recovering automatically from failures without manual intervention. With first-class Kubernetes and Python SDK support, Flyte enables advanced concurrency, dynamic pipelines, and integration with popular ML frameworks and app deployment tools. It's ideal for ML engineers, AI teams, and organizations needing to streamline and productionize their ML pipelines and agent-based apps.